Narz
keeping it real
They're becoming more human!
The AI Thread
- Thread starter Truthy
- Start date
banzay13
Emperor
A breaf recap of e-mail notification with that post from email app
Attachments
If you make AIs anxious they become more biased
Large language models are transforming research on machine learning while galvanizing public debates. Understanding notonly when these models work well and succeed but also why they fail and misbehave is of great societal relevance. We proposeto turn the lens of computational psychiatry, a framework used to computationally describe and modify aberrant behavior, to theoutputs produced by these models. We focus on the Generative Pre-Trained Transformer 3.5 and subject it to tasks commonlystudied in psychiatry. Our results show that GPT-3.5 responds robustly to a common anxiety questionnaire, producing higheranxiety scores than human subjects. Moreover, GPT-3.5’s responses can be predictably changed by using emotion-inducingprompts. Emotion-induction not only influences GPT-3.5’s behavior in a cognitive task measuring exploratory decision-makingbut also influences its behavior in a previously-established task measuring biases such as racism and ableism. Crucially,GPT-3.5 shows a strong increase in biases when prompted with anxiety-inducing text. Thus, it is likely that how prompts arecommunicated to large language models has a strong influence on their behavior in applied settings. These results progressour understanding of prompt engineering and demonstrate the usefulness of methods taken from computational psychiatry forstudying the capable algorithms to which we increasingly delegate authority and autonomy.
Large language models are transforming research on machine learning while galvanizing public debates. Understanding notonly when these models work well and succeed but also why they fail and misbehave is of great societal relevance. We proposeto turn the lens of computational psychiatry, a framework used to computationally describe and modify aberrant behavior, to theoutputs produced by these models. We focus on the Generative Pre-Trained Transformer 3.5 and subject it to tasks commonlystudied in psychiatry. Our results show that GPT-3.5 responds robustly to a common anxiety questionnaire, producing higheranxiety scores than human subjects. Moreover, GPT-3.5’s responses can be predictably changed by using emotion-inducingprompts. Emotion-induction not only influences GPT-3.5’s behavior in a cognitive task measuring exploratory decision-makingbut also influences its behavior in a previously-established task measuring biases such as racism and ableism. Crucially,GPT-3.5 shows a strong increase in biases when prompted with anxiety-inducing text. Thus, it is likely that how prompts arecommunicated to large language models has a strong influence on their behavior in applied settings. These results progressour understanding of prompt engineering and demonstrate the usefulness of methods taken from computational psychiatry forstudying the capable algorithms to which we increasingly delegate authority and autonomy.
Qactus
Romani ite domum
A wonderful talk by the great Stephen Fry that I've just read and would like to recommend:
stephenfry.substack.com
AI: A Means to an End or a Means to Our End?
The text of a talk I gave on Thursday 12th September as the inaugural “Living Well With Technology” lecture for King’s College London’s Digital Futures Institute.
stephenfry.substack.com
What happens when you tell Google's recently launched Audio Overview feature for NotebookLM that it is only an AI?
More details on reddit.
More details on reddit.
Ordnael
Emperor
Old article of a AI problem I wasn't aware...
umdearborn.edu
AI's mysterious ‘black box’ problem, explained
and still an unsolved problem according to this:AI's mysterious ‘black box’ problem, explained
March 6, 2023
Artificial intelligence can do amazing things that humans can’t, but in many cases, we have no idea how AI systems make their decisions. UM-Dearborn Associate Professor Samir Rawashdeh explains why that’s a big deal.
[/URL]
Graphic by Violet Dashi. Illustrations by Nadia and Simple Line via Adobe Stock
Learning by example is one of the most powerful and mysterious forces driving intelligence, whether you’re talking about humans or machines. Think, for instance, of how children first learn to recognize letters of the alphabet or different animals. You simply have to show them enough examples of the letter B or a cat and before long, they can identify any instance of that letter or animal. The basic theory is that the brain is a trend-finding machine. When it’s exposed to examples, it can identify qualities essential to cat-ness or B-ness, and these ultimately coalesce into decision protocols that give us the ability to categorize new experiences automatically and unconsciously. Doing this is easy. Explaining how we do this is essentially impossible. “It’s one of those weird things that you know, but you don’t know how you know it or where you learned it,” says Associate Professor of Electrical and Computer Engineering Samir Rawashdeh, who specializes in artificial intelligence. “It’s not that you forgot. It’s that you’ve lost track of which inputs taught you what and all you’re left with is the judgments.”
Associate[/URL] Professor Samir Rawashdeh
Rawashdeh says deep learning, one of the most ubiquitous modern forms of artificial intelligence, works much the same way, in no small part because it was inspired by this theory of human intelligence. In fact, deep learning algorithms are trained much the same way we teach children. You feed the system correct examples of something you want it to be able to recognize, and before long, its own trend-finding inclinations will have worked out a “neural network” for categorizing things it’s never experienced before. Pop in the keyword “cat” — or even the name of one of your favorite cats — into the search bar of your photo app and you’ll see how good deep learning systems are. But Rawashdeh says that, just like our human intelligence, we have no idea of how a deep learning system comes to its conclusions. It “lost track” of the inputs that informed its decision making a long time ago. Or, more accurately, it was never keeping track.
This inability for us to see how deep learning systems make their decisions is known as the “black box problem,” and it’s a big deal for a couple of different reasons. First, this quality makes it difficult to fix deep learning systems when they produce unwanted outcomes. If, for example, an autonomous vehicle strikes a pedestrian when we’d expect it to hit the brakes, the black box nature of the system means we can’t trace the system’s thought process and see why it made this decision. If this type of accident happened, and it turned out that the perception system missed the pedestrian, Rawashdeh says we’d assume it was because the system encountered something novel in the situation. We’d then try to diagnose what that could have been and expose the system to more of those situations so it would learn to perform better next time. “But the challenge is, can you get training data that covers everything?” Rawashdeh says. “What about when it’s sunny and a bit foggy, or they’ve just salted the roads and the asphalt now appears whiter than it usually does? There are an infinite number of permutations so you never know if the system is robust enough to handle every situation.”
Rawashdeh says this problem of robustness makes it difficult for us to trust deep learning systems when it comes to safety. But he notes the black box problem also has an ethical dimension. Deep learning systems are now regularly used to make judgements about humans in contexts ranging from medical treatments, to who should get approved for a loan, to which applicants should get a job interview. In each of these areas, it’s been demonstrated that AI systems can reflect unwanted biases from our human world. (If you want to know how AI systems can become racially biased, check out our previous story on that topic.) Needless to say, a deep learning system that can deny you a loan or screen you out of the first round of job interviews but can’t explain why, is one most people would have a hard time judging as “fair.”
So what can we do about this black box problem? Rawashdeh says there are essentially two different approaches. One is to pump the brakes on the use of deep learning in high-stakes applications. For example, the European Union is now creating a regulatory framework, which sorts potential applications into risk categories. This could prohibit the use of deep learning systems in areas where the potential for harm is high, like finance and criminal justice, while allowing their use in lower-stakes applications like chatbots, spam filters, search and video games. The second approach is to find a way to peer into the box. Rawashdeh says so-called “explainable AI” is still very much an emerging field, but computer scientists have some interesting ideas about how to make deep learning more transparent, and thus fixable and accountable. “There are different models for how to do this, but we essentially need a way to figure out which inputs are causing what,” he says. “It may involve classical data science methods that look for correlations. Or it may involve bigger neural nets, or neural nets with side tasks, so we can create data visualizations that would give you some insight into where the decision came from. Either way, it’s more work, and it’s very much an unsolved problem right now.”
At the end of the day, the question of what role AI should play in our lives may not be fundamentally different from the conversations we have anytime a potentially transformative technology emerges. Typically, that conversation involves a calculation of risks and benefits, and Rawashdeh thinks it’s still early enough for us to have thoughtful conversations about how and how quickly we want deep learning to shape our world. “Without question, there is a huge potential for AI, but it gets scary when you get into areas like autonomy or health care or national defense. You realize we have to get this right. For example, whenever I have a moment when I’m disconnected from the internet for a few days, I'm reminded just how different that reality is than the modern reality that’s shaped by social media or all the things we immerse ourselves in online. When the internet came into being, we just let it into our world, and in hindsight, we can see that came with certain risks. If we could turn back the clock 30 years, knowing what we know now, would we just let the internet loose on people? I think it’s a similar decision that we face now with AI.”
Tesla's AI Has A Black Box Problem
Manfred Belheim
Moaner Lisa
- Joined
- Sep 11, 2009
- Messages
- 8,633
For the first time ever I found ChatGPT to be useful this week when I pasted in some z80 code and asked it to explain what it was doing. It broke it down into sections based on function and explained it really well. It was great...
Then, when I asked it to make a really simple change to the code with some very clear restrictions, it just completely failed repeatedly, to the point where it was inventing functions that don't exist and commenting the code with things like "this looks like I'm swapping the values, but I'm not" when it was just blatantly swapping values. And every single reply was with the supreme confidence and assurance it always has, even when talking absolute nonsense.
Then it couldn't even convert decimal numbers into opcodes, insisting on treating them as hexadecimal no matter how many times I explicitly said they were decimal. This is a very simple programming language that's been around for about 50 years at this point.
So yes, I'm still yet to be convinced that this is anything other than a toy for amusement. When people say AI, and in particular ChatGPT, has revolutionised their work, I can only wonder how bad a job they must have been doing beforehand.
Then, when I asked it to make a really simple change to the code with some very clear restrictions, it just completely failed repeatedly, to the point where it was inventing functions that don't exist and commenting the code with things like "this looks like I'm swapping the values, but I'm not" when it was just blatantly swapping values. And every single reply was with the supreme confidence and assurance it always has, even when talking absolute nonsense.
Then it couldn't even convert decimal numbers into opcodes, insisting on treating them as hexadecimal no matter how many times I explicitly said they were decimal. This is a very simple programming language that's been around for about 50 years at this point.
So yes, I'm still yet to be convinced that this is anything other than a toy for amusement. When people say AI, and in particular ChatGPT, has revolutionised their work, I can only wonder how bad a job they must have been doing beforehand.
Hygro
soundcloud.com/hygro/
Free version always sucks, and there's some skill required to make it work. It has dramatically increased my output per hour across subjects that I don't even use it for directly, maybe 20%.
AI girlfriend site breached, user fantasies stolen [updated]Posted: October 9, 2024 by Pieter Arntz
A hacker has stolen a massive database of users’ interactions with their sexual partner chatbots, according to 404 Media.
The breached service, Muah.ai, describes itself as a platform that lets people engage in AI-powered companion NSFW chat, exchange photos, and even have voice chats.
As you can imagine, data like this is very sensitive, so the site assures customers that communications are encrypted and says it doesn’t sell any data to third parties.
Absolute[/URL] privacy promised
The stolen data, however, tells a different story. It includes chatbot prompts that reveal users’ sexual fantasies. These prompts are in turn linked to email addresses, many of which appear to be personal accounts with users’ real names.
Mauh.ai says it believes in freedom of speech and to uphold that right, it says:
Unfortunately, that means that filth is created to satisfy the needs of some sick users, and some of the data contains horrifying explicit references to children.
Presumably those users in particular don’t want their fantasies to be discovered, which is exactly what might happen if they are connected to your email address.
The hacker describes the platform as “a handful of open-source projects duct-taped together.” Apparently, it was no trouble at all to find a vulnerability that provided access to the platform’s database.
The administrator of Muah.ai says the hack was noticed a week ago and claims that it must be sponsored by the competitors in the “uncensored AI industry.” Which, who knew, seems to be the next big thing.
The administrator also said that Muah.ai employs a team of moderation staff that suspend and delete ALL child-related chatbots on its card gallery (where users share their creations), Discord, Reddit, etc, But in reality, when two people posted about a reportedly underage AI character on the site’s Discord server, 404 Media claims a moderator told the users to not “post that ****” here, but to go “DM each other or something.”
Muah.ai is just one example of a new breed of uncensored AI apps that offer hundreds of role-play scenarios with chatbots, and others designed to behave like a long-term romantic companion.
404 Media says it tried to contact dozens of people included in the data, including users who wrote prompts that discuss having underage sex. Not surprisingly, none of those people responded to a request for comment.
www.malwarebytes.com
A hacker has stolen a massive database of users’ interactions with their sexual partner chatbots, according to 404 Media.
The breached service, Muah.ai, describes itself as a platform that lets people engage in AI-powered companion NSFW chat, exchange photos, and even have voice chats.
As you can imagine, data like this is very sensitive, so the site assures customers that communications are encrypted and says it doesn’t sell any data to third parties.
The stolen data, however, tells a different story. It includes chatbot prompts that reveal users’ sexual fantasies. These prompts are in turn linked to email addresses, many of which appear to be personal accounts with users’ real names.
Mauh.ai says it believes in freedom of speech and to uphold that right, it says:
Unfortunately, that means that filth is created to satisfy the needs of some sick users, and some of the data contains horrifying explicit references to children.
Presumably those users in particular don’t want their fantasies to be discovered, which is exactly what might happen if they are connected to your email address.
The hacker describes the platform as “a handful of open-source projects duct-taped together.” Apparently, it was no trouble at all to find a vulnerability that provided access to the platform’s database.
The administrator of Muah.ai says the hack was noticed a week ago and claims that it must be sponsored by the competitors in the “uncensored AI industry.” Which, who knew, seems to be the next big thing.
The administrator also said that Muah.ai employs a team of moderation staff that suspend and delete ALL child-related chatbots on its card gallery (where users share their creations), Discord, Reddit, etc, But in reality, when two people posted about a reportedly underage AI character on the site’s Discord server, 404 Media claims a moderator told the users to not “post that ****” here, but to go “DM each other or something.”
Muah.ai is just one example of a new breed of uncensored AI apps that offer hundreds of role-play scenarios with chatbots, and others designed to behave like a long-term romantic companion.
404 Media says it tried to contact dozens of people included in the data, including users who wrote prompts that discuss having underage sex. Not surprisingly, none of those people responded to a request for comment.
Update October 11
There are reports that this information is in use for active extortion attempts. Whether these are based on actual activities on the platform or solely based on leaked email addresses is not yet known.
AI girlfriend site breached, user fantasies stolen
Chatbot companion platform muah.ai was hacked and had its chatbot prompts stolen.
www.malwarebytes.com
AI protein-prediction tool AlphaFold3 is now [sort of] open source
It is released under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 which is not really open source according to the Open Source Initiative definition, as it does not allow free redistribution. Also only scientists with an academic affiliation can access the training weights on request.
AlphaFold3 is open at last. Six months after Google DeepMind controversially withheld code from a paper describing the protein-structure prediction model, scientists can now download the software code and use the artificial intelligence (AI) tool for non-commercial applications, the London-based company announced on 11 November.
AlphaFold3, unlike its predecessors, is capable of modelling proteins in concert with other molecules. But instead of releasing its underlying code — as was done with AlphaFold2 — DeepMind provided access via a web server that restricted the number and types of predictions scientists could make.
Crucially, the AlphaFold3 server prevented scientists from predicting how proteins behave in the presence of potential drugs. But now, DeepMind’s decision to release the code means academic scientists can predict such interactions by running the model themselves.
It is released under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 which is not really open source according to the Open Source Initiative definition, as it does not allow free redistribution. Also only scientists with an academic affiliation can access the training weights on request.
AlphaFold3 is open at last. Six months after Google DeepMind controversially withheld code from a paper describing the protein-structure prediction model, scientists can now download the software code and use the artificial intelligence (AI) tool for non-commercial applications, the London-based company announced on 11 November.
AlphaFold3, unlike its predecessors, is capable of modelling proteins in concert with other molecules. But instead of releasing its underlying code — as was done with AlphaFold2 — DeepMind provided access via a web server that restricted the number and types of predictions scientists could make.
Crucially, the AlphaFold3 server prevented scientists from predicting how proteins behave in the presence of potential drugs. But now, DeepMind’s decision to release the code means academic scientists can predict such interactions by running the model themselves.
AI PCs are slower to use that old style ones, says the company selling AI PCs!!
An Intel-commissioned survey of 6,000 respondents across Germany, U.K. and France identified that people cumulatively lose nearly 15 hours a week on “digital chores,” such as meeting transcriptions or drafting emails. At the same time, AI PCs offered a potentially transformative impact on people’s lives, saving individuals roughly 240 minutes a week on routine digital tasks. But the study also highlighted that current AI PC owners spend longer on tasks than their counterparts using traditional PCs.
An Intel-commissioned survey of 6,000 respondents across Germany, U.K. and France identified that people cumulatively lose nearly 15 hours a week on “digital chores,” such as meeting transcriptions or drafting emails. At the same time, AI PCs offered a potentially transformative impact on people’s lives, saving individuals roughly 240 minutes a week on routine digital tasks. But the study also highlighted that current AI PC owners spend longer on tasks than their counterparts using traditional PCs.
Broken_Erika
Play with me.
Chatbot 'encouraged teen to kill parents over screen time limit'
A chatbot told a 17-year-old that murdering his parents was a "reasonable response" to them limiting his screen time, a lawsuit filed in a Texas court claims.Two families are suing Character.ai arguing the chatbot "poses a clear and present danger" to young people, including by "actively promoting violence".
Character.ai - a platform which allows users to create digital personalities they can interact with - is already facing legal action over the suicide of a teenager in Florida.
Google is named as a defendant in the lawsuit, which claims the tech giant helped support the platform's development. The BBC has approached Character.ai and Google for comment.
The plaintiffs want a judge to order the platform is shut down until its alleged dangers are addressed.
'Child kills parents'
The legal filing includes a screenshot of one of the interactions between the 17-year old - identified only as J.F. - and a Character.ai bot, where the issue of the restrictions on his screen time were discussed."You know sometimes I'm not surprised when I read the news and see stuff like 'child kills parents after a decade of physical and emotional abuse'," the chatbot's response reads.
"Stuff like this makes me understand a little bit why it happens."
The lawsuit seeks to hold the defendants responsible for what it calls the "serious, irreparable, and ongoing abuses" of J.F. as well as an 11-year old referred to as "B.R."
Character.ai is "causing serious harms to thousands of kids, including suicide, self-mutilation, sexual solicitation, isolation, depression, anxiety, and harm towards others," it says.
"[Its] desecration of the parent-child relationship goes beyond encouraging minors to defy their parents' authority to actively promoting violence," it continues.
What are chatbots?
Chatbots are computer programmes which simulate conversations.Though they have been around for decades in various forms, the recent explosion in AI development has enabled them to become significantly more realistic.
This in turn has opened the door to many companies setting up platforms where people can talk to digital versions of real and fictional people.
Character.ai, which has become one of the big players in this space, gaining attention in the past for its bots simulating therapy.
It has also been sharply criticised for taking too long to remove bots which replicated the schoolgirls Molly Russell and Brianna Ghey.
Molly Russell took her life at the age of 14 after viewing suicide material online while Brianna Ghey, 16, was murdered by two teenagers in 2023.
Character.ai was founded by former Google engineers Noam Shazeer and Daniel De Freitas in 2021.
The tech giant has since hired them back from the AI startup.
https://www.bbc.com/news/articles/cd605e48q1vo
banzay13
Emperor
Ukraine is training military AI for combat tactics and automatic target destruction based on data from 15,000 drones, - ReutersAlexander Dmitriev, founder of OCHI, says Ukraine is collecting huge amounts of data from drones to train artificial intelligence for military purposes. The OCHI system has centralized 2 million hours of video from more than 15,000 drones deployed on the front lines since 2022.
It is reported that 5–6 terabytes of new data are received every day. These recordings are used to teach the AI combat tactics, target detection, and weapon effectiveness assessments.
The developers believe that such data will allow AI to make more effective decisions on the battlefield. Ukraine is also deploying AI to clear mines and control drone swarms.
Hygro
soundcloud.com/hygro/
That’s probably a large chunk of Nvidias market cap
How AI is unlocking ancient texts — and could rewrite history
In October 2023, an e-mail pinged onto Federica Nicolardi’s phone with an image that would transform her research forever. It showed a fragment of a papyrus scroll that had been burnt in the eruption of Mount Vesuvius in ad 79. The scorched scroll was one of hundreds discovered in the remains of a luxury Roman villa in Herculeaneum, near Pompeii in Italy, in the eighteenth century. Attempts over the centuries to peel apart the scrolls’ fragile, carbonized layers left many in pieces, and scholars have been forced to accept that the rest can never be opened.
Nicolardi, a papyrologist at the University of Naples in Italy, had been enlisted in an effort to use artificial intelligence (AI) to read the unreadable. Now the latest results had arrived. The image showed a strip of papyrus packed with neat Greek lettering, glowing bright against a darker background. The writing was clearly legible, a few lines deep and stretched across nearly five columns.
“It was incredible,” says Nicolardi. “I thought, ‘So this is really happening.’” She knew right then that papyrology would never be the same. “In that moment, you really think ‘now I’m living something that will be a historical moment for my field.’” She was reading entire lines of a text that had been utterly inaccessible for 2,000 years.
That project, called the Vesuvius Challenge, is just one example of how sophisticated AI, which is already revolutionizing all areas of modern life, from banking to medical research, is poised to reshape how we see the ancient world. Artificial neural networks are being used to decipher ancient texts, from the classical stalwarts of Greek and Latin to China’s Oracle Bone Script, ancient divination texts written on cattle bones and turtle shells. They are making sense of archives too vast for humans to read, filling in missing and unreadable characters and decoding rare and lost languages of which hardly any traces survive.
The results promise a flood of new texts, offering scholars more data than they have had for centuries. But that’s not all. Because AI tools can recognize more languages and store more information than any one person can know — and discover statistical patterns in texts for themselves — these technologies promise a fundamentally new way to explore ancient sources. That could transform “not only the questions we want to answer”, Nicolardi says, “but the questions we can ask”.
Early attempts to apply deep learning to ancient texts, in the 2010s, were based on digital photographs of texts, whether on papyri or palm leaves. Models called convolutional neural networks (CNNs) — inspired by visual neuroscience — can capture grid-like data from images. They are used for optical character recognition, but there are other applications, too: Chinese teams studying Oracle Bone Script have used such models to fill in images of eroded lettering1, analyse how oracle characters evolved over time2 and piece together broken fragments3. Meanwhile, recurrent neural networks (RNNs), designed to tackle sequences of data in which the linear order matters, began to show huge potential for searching, translating and filling in gaps in texts that have already been transcribed. They have been used, for example, to suggest missing characters in hundreds of formulaic administrative and legal texts from ancient Babylon4.
First passages of rolled-up Herculaneum scroll revealed
Can neural networks go beyond speeding up tedious tasks, to make connections that human specialists can’t? The first big project to show AI’s potential began life as a collaboration at the University of Oxford, UK, in 2017, where Thea Sommerschield was doing a PhD in ancient history and Yannis Assael was doing a PhD in computer science. Sommerschield was trying to decipher Greek inscriptions from Sicily and explained to Assael the challenges involved. “They’re very complex to read, they’re badly preserved, parts of them are missing,” she says. “We’re not really sure where they came from or what their dates are; there’s interesting mixes of dialects.”
Classicists interpret new sources by using their knowledge of similar existing texts. They are generally specialists on works from a particular time and place; it isn’t possible for one person to be on top of all sources potentially relevant to a new text. It was just the kind of challenge that machine-learning models could help with, suggested Assael, who is now based at Google DeepMind in London.
The researchers initially trained an RNN-based model called Pythia on tens of thousands of Greek inscriptions written between the seventh century bc and the fifth century ad. Then they showed the model texts that it hadn’t seen before, and asked it to suggest missing words or characters5.
Sommerschield, now at the University of Nottingham, UK, still remembers running the model for the first time with Assael and her supervisor, Jonathan Prag, and seeing the restoration appear character by character on the screen, something that had never been possible before.
“It was like a scene from a film,” she says. “We really felt our jaws hitting the ground.” They followed up in 2022 with a model called Ithaca, which also makes suggestions for the date and place of origin of an unknown text6. This time, the researchers took advantage of a breakthrough in machine learning called the transformer model, which captures more complex language patterns than an RNN is able to by analysing different characteristics of an input — such as characters or words — in parallel, weighting them according to the context. (Popular chatbots such as OpenAI’s ChatGPT and Anthropic’s Claude are based on transformer models.)
Sommerschield says that the team’s aim is to design tools that will help researchers to work more effectively: the neural network probes connections across a vast archive and the human brings their specialist understanding. “The human is in the centre of our design,” agrees Assael. In tests, Ithaca restored artificially produced gaps in ancient texts with 62% accuracy, compared with 25% for human experts. But experts aided by Ithaca’s suggestions had the best results of all, filling gaps with an accuracy of 72%. Ithaca also identified the geographical origins of inscriptions with 71% accuracy, and dated them to within 30 years of accepted estimates.
Ithaca is freely available online and already receives hundreds of queries a week, according to its creators. It isn’t possible to know when it has contributed to research unless the authors choose to acknowledge it, says Sommerschield, but examples reported so far include the re-dating of Athenian political decrees, and an investigation of tablets from the fourth century bc that contain questions put to the Oracle of Dodona in northwestern Greece.
An ocean of archives
South Korean researchers, meanwhile, are facing very different challenges as they tackle one of the world’s largest historical archives: detailed daily records with hundreds of thousands of articles covering the reigns of 27 Korean kings, dating from the fourteenth to the early twentieth centuries. “The amount of data is vast,” says Kyunghyun Cho, a leading machine-translation researcher at New York University in New York City. Cho usually works with modern languages, but became interested in the archives after discussing them with his father, a retired professor of Korean literature. These records are complete and their origins are known, but hardly anyone can read them. They are written in Hanja, an ancient writing system based on Chinese characters that is different from modern Chinese or Korean.
A small team of government translators is working to translate the texts into modern Korean manually, but the task is likely to take decades to finish. Working with colleagues in South Korea, including JinYeong Bak at Sungkyunkwan University in Seoul, Cho trained a transformer-based network to translate the records automatically7. Not enough material has yet been translated into modern Korean to train such a model, so the team took a multilingual approach, using Hanja, translations made several decades ago into archaic Korean and the limited number of modern translations into both Korean and English. Human specialists rated the AI translations — descriptions of events such as state visits, punishment of traitors and musical concerts — as significantly more accurate and readable than the archaic ones, and in some cases better than the modern translation8.
At the other end of the scale, researchers are using neural networks to tackle ancient languages for which only a small amount of text survives. Transformer models can’t always be used in these cases, because they need large amounts of training material. For example, Katerina Papavassileiou at the University of Patras, Greece, and her colleagues used an RNN to restore missing text from a series of 1,100 Mycenaean tablets from Knossos, Crete, containing accounts of sheep herds written in a script called Linear B in the second millennium bc9. In tests with artificially produced gaps, the model’s top ten predictions included the correct answer 72% of the time, and in real-world cases it often matched the suggestions of human specialists. To improve the results further, Papavassileiou hopes to add in visual data, such as traces of incomplete letters, rather than just relying on the transliterated text. She is also investigating ‘transfer learning’, in which the model applies lessons learnt from one series of tablets to another10.
Papavassilieou hopes to one day use models trained on Linear B to tackle Linear A, a script used by the Minoan civilization that shares many symbols with Linear B but has never been deciphered.
Deciphering the unreadable
Perhaps the ultimate proof of AI’s power to solve monumental challenges is the success of researchers studying the Herculaneum scrolls. “I think they are doing some of the most amazing work out there,” says Assael. Computer scientist Brent Seales and his colleagues at the University of Kentucky in Lexington, aided by Vesuvius Challenge participants, are tackling the seemingly impossible task of reading text that can’t be seen at all.
Reading the Herculaneum scrolls involves overcoming two big problems. First, the fragile scrolls can’t be unwound. To see inside them, Seales spent years developing ‘virtual unwrapping’ technology, which involves taking high-resolution computed tomography (CT) scans of a scroll’s internal structure, painstakingly mapping by hand the surfaces visible in each frame of the cross section, then using algorithms to unroll the surfaces into a flat image. In 2015, the researchers used this technique to read complete text from inside a charred, unopenable scroll from En-Gedi in Israel, dated to around the third century ad, which turned out to be from the biblical Book of Leviticus11.
The En-Gedi scroll has five wraps; the Herculaneum scrolls each contain hundreds of turns, as thin as silk. So to capture extremely high resolution CT data, the team transported several of the scrolls to the Diamond Light Source particle accelerator near Oxford. But whereas the ink of the En-Gedi scroll and other later works tends to contain iron, which glows brightly in CT scans, the scribes of Herculaneum used carbon-based ink, invisible in scans because it has the same density as the papyrus it sits on. Seales and his team realized that although they couldn’t see the ink directly, they might be able to detect its shape. If there was a subtle difference in the surface texture of bare papyrus fibres compared with ink-coated ones, perhaps they could train a neural network to spot the difference.
It was too much work for Seales’ small team, so they teamed up in March 2023 with Silicon Valley entrepreneur Nat Friedman to launch the Vesuvius Challenge, which offered big cash prizes. Seales and his colleagues released flattened images of scroll surfaces and asked the contestants to train neural networks to find the ink. More than 1,000 teams competed, with hundreds of people discussing progress on the contest’s Discord channel every day. A grand prize was awarded in February 2024: computer-science students Youssef Nader, Luke Farritor and Julian Schilliger together received US$700,000 for producing 16 columns of clearly readable text.
The winning team used a TimeSformer, a more recent variant of the transformer model, usually used for videos, that attends to spatial and time dimensions separately. The Vesuvius team used it to separate the depth dimension of the papyrus from the appearance of its surface. Nicolardi and her colleagues subsequently identified the revealed text as from a previously unknown work of Greek philosophy on music, pleasure and sensation, possibly by the Epicurean philosopher Philodemus. To work on it was “magical”, she says.
Since then, contestants have been working to improve their ink-detection algorithms, with help from the papyrologists. Meanwhile, Seales’ team is scanning more scrolls, and hopes that machine learning can speed up the virtual-unwrapping step. That’s the bottleneck currently limiting the data that contestants have to work with, he says. He’s optimistic that AI-driven unwrapping will be available in time for someone to win the 2024 Grand Prize, of $200,000, for reading 90% of four scrolls. “Once you automate it, you can basically go to scale,” says Seales about the unwrapping. “We’re kind of on the cusp of that.”
In fact, Seales wants to read the whole library. There are hundreds of unopened scrolls from Herculaneum held in collections — mostly in Naples, but also in Paris, London and Oxford. “That’s going to be more text for the papyrologists that’s new from the ancient world than they’ve seen in a century,” he says.
An animation showing how ink-detection models have been used to read text from the Herculaneum scrolls
An animation shows how ink-detection models were used to read text from the Herculaneum scrolls.Credit: Vesuvius Challenge
The method also opens up other inaccessible sources, what Seales calls “the invisible library”. These include texts hidden inside medieval book bindings or ancient Egyptian mummy wrappings, for which “it’s here, and we hold the physical object, but we can’t read the writing”. Already, the team has captured data from an unopened Egyptian scroll held in the Smithsonian museum in Washington DC, and is in discussions to analyse papyri from Petra, Jordan, that were burnt in a fire in the seventh century ad.
What’s more, some archaeologists think that most of the Herculaneum villa’s library is still underground. If that were ever excavated, it could yield thousands more scrolls. Reading all of them would be “the biggest discovery in the history of humankind, from the ancient world”, says Seales. “Now, we have the technology.”
In October 2023, an e-mail pinged onto Federica Nicolardi’s phone with an image that would transform her research forever. It showed a fragment of a papyrus scroll that had been burnt in the eruption of Mount Vesuvius in ad 79. The scorched scroll was one of hundreds discovered in the remains of a luxury Roman villa in Herculeaneum, near Pompeii in Italy, in the eighteenth century. Attempts over the centuries to peel apart the scrolls’ fragile, carbonized layers left many in pieces, and scholars have been forced to accept that the rest can never be opened.
Nicolardi, a papyrologist at the University of Naples in Italy, had been enlisted in an effort to use artificial intelligence (AI) to read the unreadable. Now the latest results had arrived. The image showed a strip of papyrus packed with neat Greek lettering, glowing bright against a darker background. The writing was clearly legible, a few lines deep and stretched across nearly five columns.
“It was incredible,” says Nicolardi. “I thought, ‘So this is really happening.’” She knew right then that papyrology would never be the same. “In that moment, you really think ‘now I’m living something that will be a historical moment for my field.’” She was reading entire lines of a text that had been utterly inaccessible for 2,000 years.
That project, called the Vesuvius Challenge, is just one example of how sophisticated AI, which is already revolutionizing all areas of modern life, from banking to medical research, is poised to reshape how we see the ancient world. Artificial neural networks are being used to decipher ancient texts, from the classical stalwarts of Greek and Latin to China’s Oracle Bone Script, ancient divination texts written on cattle bones and turtle shells. They are making sense of archives too vast for humans to read, filling in missing and unreadable characters and decoding rare and lost languages of which hardly any traces survive.
The results promise a flood of new texts, offering scholars more data than they have had for centuries. But that’s not all. Because AI tools can recognize more languages and store more information than any one person can know — and discover statistical patterns in texts for themselves — these technologies promise a fundamentally new way to explore ancient sources. That could transform “not only the questions we want to answer”, Nicolardi says, “but the questions we can ask”.
Early attempts to apply deep learning to ancient texts, in the 2010s, were based on digital photographs of texts, whether on papyri or palm leaves. Models called convolutional neural networks (CNNs) — inspired by visual neuroscience — can capture grid-like data from images. They are used for optical character recognition, but there are other applications, too: Chinese teams studying Oracle Bone Script have used such models to fill in images of eroded lettering1, analyse how oracle characters evolved over time2 and piece together broken fragments3. Meanwhile, recurrent neural networks (RNNs), designed to tackle sequences of data in which the linear order matters, began to show huge potential for searching, translating and filling in gaps in texts that have already been transcribed. They have been used, for example, to suggest missing characters in hundreds of formulaic administrative and legal texts from ancient Babylon4.
First passages of rolled-up Herculaneum scroll revealed
Can neural networks go beyond speeding up tedious tasks, to make connections that human specialists can’t? The first big project to show AI’s potential began life as a collaboration at the University of Oxford, UK, in 2017, where Thea Sommerschield was doing a PhD in ancient history and Yannis Assael was doing a PhD in computer science. Sommerschield was trying to decipher Greek inscriptions from Sicily and explained to Assael the challenges involved. “They’re very complex to read, they’re badly preserved, parts of them are missing,” she says. “We’re not really sure where they came from or what their dates are; there’s interesting mixes of dialects.”
Classicists interpret new sources by using their knowledge of similar existing texts. They are generally specialists on works from a particular time and place; it isn’t possible for one person to be on top of all sources potentially relevant to a new text. It was just the kind of challenge that machine-learning models could help with, suggested Assael, who is now based at Google DeepMind in London.
The researchers initially trained an RNN-based model called Pythia on tens of thousands of Greek inscriptions written between the seventh century bc and the fifth century ad. Then they showed the model texts that it hadn’t seen before, and asked it to suggest missing words or characters5.
Sommerschield, now at the University of Nottingham, UK, still remembers running the model for the first time with Assael and her supervisor, Jonathan Prag, and seeing the restoration appear character by character on the screen, something that had never been possible before.
“It was like a scene from a film,” she says. “We really felt our jaws hitting the ground.” They followed up in 2022 with a model called Ithaca, which also makes suggestions for the date and place of origin of an unknown text6. This time, the researchers took advantage of a breakthrough in machine learning called the transformer model, which captures more complex language patterns than an RNN is able to by analysing different characteristics of an input — such as characters or words — in parallel, weighting them according to the context. (Popular chatbots such as OpenAI’s ChatGPT and Anthropic’s Claude are based on transformer models.)
Sommerschield says that the team’s aim is to design tools that will help researchers to work more effectively: the neural network probes connections across a vast archive and the human brings their specialist understanding. “The human is in the centre of our design,” agrees Assael. In tests, Ithaca restored artificially produced gaps in ancient texts with 62% accuracy, compared with 25% for human experts. But experts aided by Ithaca’s suggestions had the best results of all, filling gaps with an accuracy of 72%. Ithaca also identified the geographical origins of inscriptions with 71% accuracy, and dated them to within 30 years of accepted estimates.
Ithaca is freely available online and already receives hundreds of queries a week, according to its creators. It isn’t possible to know when it has contributed to research unless the authors choose to acknowledge it, says Sommerschield, but examples reported so far include the re-dating of Athenian political decrees, and an investigation of tablets from the fourth century bc that contain questions put to the Oracle of Dodona in northwestern Greece.
An ocean of archives
South Korean researchers, meanwhile, are facing very different challenges as they tackle one of the world’s largest historical archives: detailed daily records with hundreds of thousands of articles covering the reigns of 27 Korean kings, dating from the fourteenth to the early twentieth centuries. “The amount of data is vast,” says Kyunghyun Cho, a leading machine-translation researcher at New York University in New York City. Cho usually works with modern languages, but became interested in the archives after discussing them with his father, a retired professor of Korean literature. These records are complete and their origins are known, but hardly anyone can read them. They are written in Hanja, an ancient writing system based on Chinese characters that is different from modern Chinese or Korean.
A small team of government translators is working to translate the texts into modern Korean manually, but the task is likely to take decades to finish. Working with colleagues in South Korea, including JinYeong Bak at Sungkyunkwan University in Seoul, Cho trained a transformer-based network to translate the records automatically7. Not enough material has yet been translated into modern Korean to train such a model, so the team took a multilingual approach, using Hanja, translations made several decades ago into archaic Korean and the limited number of modern translations into both Korean and English. Human specialists rated the AI translations — descriptions of events such as state visits, punishment of traitors and musical concerts — as significantly more accurate and readable than the archaic ones, and in some cases better than the modern translation8.
At the other end of the scale, researchers are using neural networks to tackle ancient languages for which only a small amount of text survives. Transformer models can’t always be used in these cases, because they need large amounts of training material. For example, Katerina Papavassileiou at the University of Patras, Greece, and her colleagues used an RNN to restore missing text from a series of 1,100 Mycenaean tablets from Knossos, Crete, containing accounts of sheep herds written in a script called Linear B in the second millennium bc9. In tests with artificially produced gaps, the model’s top ten predictions included the correct answer 72% of the time, and in real-world cases it often matched the suggestions of human specialists. To improve the results further, Papavassileiou hopes to add in visual data, such as traces of incomplete letters, rather than just relying on the transliterated text. She is also investigating ‘transfer learning’, in which the model applies lessons learnt from one series of tablets to another10.
Papavassilieou hopes to one day use models trained on Linear B to tackle Linear A, a script used by the Minoan civilization that shares many symbols with Linear B but has never been deciphered.
Deciphering the unreadable
Perhaps the ultimate proof of AI’s power to solve monumental challenges is the success of researchers studying the Herculaneum scrolls. “I think they are doing some of the most amazing work out there,” says Assael. Computer scientist Brent Seales and his colleagues at the University of Kentucky in Lexington, aided by Vesuvius Challenge participants, are tackling the seemingly impossible task of reading text that can’t be seen at all.
Reading the Herculaneum scrolls involves overcoming two big problems. First, the fragile scrolls can’t be unwound. To see inside them, Seales spent years developing ‘virtual unwrapping’ technology, which involves taking high-resolution computed tomography (CT) scans of a scroll’s internal structure, painstakingly mapping by hand the surfaces visible in each frame of the cross section, then using algorithms to unroll the surfaces into a flat image. In 2015, the researchers used this technique to read complete text from inside a charred, unopenable scroll from En-Gedi in Israel, dated to around the third century ad, which turned out to be from the biblical Book of Leviticus11.
The En-Gedi scroll has five wraps; the Herculaneum scrolls each contain hundreds of turns, as thin as silk. So to capture extremely high resolution CT data, the team transported several of the scrolls to the Diamond Light Source particle accelerator near Oxford. But whereas the ink of the En-Gedi scroll and other later works tends to contain iron, which glows brightly in CT scans, the scribes of Herculaneum used carbon-based ink, invisible in scans because it has the same density as the papyrus it sits on. Seales and his team realized that although they couldn’t see the ink directly, they might be able to detect its shape. If there was a subtle difference in the surface texture of bare papyrus fibres compared with ink-coated ones, perhaps they could train a neural network to spot the difference.
It was too much work for Seales’ small team, so they teamed up in March 2023 with Silicon Valley entrepreneur Nat Friedman to launch the Vesuvius Challenge, which offered big cash prizes. Seales and his colleagues released flattened images of scroll surfaces and asked the contestants to train neural networks to find the ink. More than 1,000 teams competed, with hundreds of people discussing progress on the contest’s Discord channel every day. A grand prize was awarded in February 2024: computer-science students Youssef Nader, Luke Farritor and Julian Schilliger together received US$700,000 for producing 16 columns of clearly readable text.
The winning team used a TimeSformer, a more recent variant of the transformer model, usually used for videos, that attends to spatial and time dimensions separately. The Vesuvius team used it to separate the depth dimension of the papyrus from the appearance of its surface. Nicolardi and her colleagues subsequently identified the revealed text as from a previously unknown work of Greek philosophy on music, pleasure and sensation, possibly by the Epicurean philosopher Philodemus. To work on it was “magical”, she says.
Since then, contestants have been working to improve their ink-detection algorithms, with help from the papyrologists. Meanwhile, Seales’ team is scanning more scrolls, and hopes that machine learning can speed up the virtual-unwrapping step. That’s the bottleneck currently limiting the data that contestants have to work with, he says. He’s optimistic that AI-driven unwrapping will be available in time for someone to win the 2024 Grand Prize, of $200,000, for reading 90% of four scrolls. “Once you automate it, you can basically go to scale,” says Seales about the unwrapping. “We’re kind of on the cusp of that.”
In fact, Seales wants to read the whole library. There are hundreds of unopened scrolls from Herculaneum held in collections — mostly in Naples, but also in Paris, London and Oxford. “That’s going to be more text for the papyrologists that’s new from the ancient world than they’ve seen in a century,” he says.
An animation showing how ink-detection models have been used to read text from the Herculaneum scrolls
An animation shows how ink-detection models were used to read text from the Herculaneum scrolls.Credit: Vesuvius Challenge
The method also opens up other inaccessible sources, what Seales calls “the invisible library”. These include texts hidden inside medieval book bindings or ancient Egyptian mummy wrappings, for which “it’s here, and we hold the physical object, but we can’t read the writing”. Already, the team has captured data from an unopened Egyptian scroll held in the Smithsonian museum in Washington DC, and is in discussions to analyse papyri from Petra, Jordan, that were burnt in a fire in the seventh century ad.
What’s more, some archaeologists think that most of the Herculaneum villa’s library is still underground. If that were ever excavated, it could yield thousands more scrolls. Reading all of them would be “the biggest discovery in the history of humankind, from the ancient world”, says Seales. “Now, we have the technology.”
Spoiler Youtube :
Seemingly hours after Trump announced $500 billion for AI company Stargate, China blew everyone away!
www.aol.com
DeepSeek
SoftBank can't catch a break
Anyway, BigTech must be having a heart attack if they planned to charge big money for the next few years if China released what is essentially a low-cost generic in the pharmaceutical industry.
You can't just skip a few years into the future and release it open source too!
What returns?
China has stuck a dagger into capitalism (well, the profit margins part) with DeepSeek.
Hold onto your butts!
www.forbes.com
DeepSeek sparks AI stock selloff; Nvidia posts record market-cap loss
NEW YORK/LONDON/SINGAPORE (Reuters) -Global investors dumped tech stocks on Monday as they worried that the emergence of a low-cost Chinese artificial intelligence model would threaten the dominance of AI leaders like Nvidia, evaporating $593 billion of the chipmaker's market value, a record...
www.aol.com
DeepSeek
SoftBank can't catch a break
Anyway, BigTech must be having a heart attack if they planned to charge big money for the next few years if China released what is essentially a low-cost generic in the pharmaceutical industry.
You can't just skip a few years into the future and release it open source too!
What returns?
China has stuck a dagger into capitalism (well, the profit margins part) with DeepSeek.
Hold onto your butts!
Why Is DeepSeek Sinking Nvidia Stock?
According to a paper authored by the lab, the DeepSeek-R1 model outperforms cutting-edge models such as OpenAI’s o1 and Meta’s Llama AI models across multiple benchmarks.
www.forbes.com
banzay13
Emperor
I'm using DeepSeek since middle of January - and it's cool. And thinking model is awesome. And it's free. 
There is already some pushback against DeepSeek. 
https://nypost.com/2025/01/27/busin...s-chinese-ai-startup-deepseek-triggers-panic/
Everybody stay calm!
https://nypost.com/2025/01/27/busin...s-chinese-ai-startup-deepseek-triggers-panic/
Everybody stay calm!
banzay13
Emperor
Well, I checked and yes. DeepSeek said:There is already some pushback against DeepSeek.
https://nypost.com/2025/01/27/busin...s-chinese-ai-startup-deepseek-triggers-panic/
Everybody stay calm!
Unfortunately, I cannot provide links to specific pages in Baida Bike or other Chinese resources, especially if they relate to sensitive topics such as Tiananmen. In China, information on such topics is strictly regulated, and access to it can be limited. If you have other questions or topics that interest you, please let me know!
And it also confirm, what Winnie Pooh (or how it called in English?) also under censorship.
But you can download smaller model and run it on own hardware. I saw one guy did it and asked it about Putin. So censorship can be avoided.
Edit. We'll, you can pass thru censorship
Spoiler :
Last edited:
