Focue Provides the Latest and Most Up-to-Date News, What You Focus On is What You Get.
⎯ 《 Focue • Com 》
ChatGPT creators try to use artificial intelligence to explain itself – and come across major problems
ChatGPT creators try to use artificial intelligence to explain itself – and come across major problems
ChatGPT’s creators have attempted to get the system to explain itself. They found that while they had some success, they ran into some issues – including the fact that artificial intelligence may be using concepts that humans do not have named for, or understanding of. Researchers at OpenAI, which developed ChatGPT, used the most recent version of its model known as GPT-4 to try and explain the behaviour of GPT-2, an earlier version. It is an attempt to overcome the so-called black box problem with large language models such as GPT. While we have a relatively good understanding of what goes into and comes out of such systems, the actual work that goes on inside remains largely mysterious. That is not only a problem because it makes things difficult for researchers. It also means that there is little way of knowing what biases might be involved in the system, or if it is providing false information to people using it, since there is no way of knowing how it came to the conclusions it did. Engineers and scientists have aimed to resolve this problem with “interpretability research”, which seeks find ways to look inside the model itself and better understand what is going on. That has often required looking at the “neutrons” that make up such a model: just like in the human brain, an AI system is made up of a host of so-called neutrons that represent parts of the data it uses. Finding those is difficult, however, since humans have had to pick through the neurons and manually inspect them to find out what they represent. But some systems have hundreds of billions of parameters and so actually getting through them all with people is impossible. Now, researchers at OpenAI have looked to use GPT-4 to automate that process, in an attempt to more quickly pick through the behaviour. They did so by attempting to create an automated process that would allow the system to provide natural language explanations of the neuron’s behaviour – and apply that to another, earlier language model. That worked in three steps: looking at the neuron in GPT-2 and having GPT-4 try and explain it, then simulating what that neuron would, and finally scoring that explanation by comparing how the simulated activation worked with the real one. Most of those explanations went badly, and GPT-4 scored itself poorly. But researchers said that they hoped the experiment showed that it would be possible to use the AI technology to explain itself, with further work. The creators came up against a range of “limitations”, however, that mean the system as it exists now is not as good as humans at explaining the behaviour. Part of the problem may be that explaining how the system is working in normal language is impossible – because the system may be using individual concepts that humans cannot name. “We focused on short natural language explanations, but neurons may have very complex behavior that is impossible to describe succinctly,” the authors write. “For example, neurons could be highly polysemantic (representing many distinct concepts) or could represent single concepts that humans don’t understand or have words for.” It also runs into problems because it is focused on specifically what each neuron does individually, and not how that might affect things later on in the text. Similarly, it can explain specific behaviour but not what mechanism is producing that behaviour, and so might spot The system also uses a lot of computing power, the researchers note. Read More Google to unveil major new AI AI robots figure out how to play football in shambolic footage White House asks hackers to break ChatGPT White House reveals plan to ‘protect’ citizens from danger of AI DeepMind boss says human-level AI is just a few years away Regulator to probe use of artificial intelligence such as ChatGPT
1970-01-01 08:00
These Students are Developing a Nose for (Fake) News
These Students are Developing a Nose for (Fake) News
LOS ANGELES--(BUSINESS WIRE)--May 10, 2023--
1970-01-01 08:00
OpenAI CEO Sam Altman to testify before Congress
OpenAI CEO Sam Altman to testify before Congress
OpenAI CEO Sam Altman will testify before Congress next Tuesday as lawmakers increasingly scrutinize the risks and benefits of artificial intelligence, according to a Senate Judiciary subcommittee.
1970-01-01 08:00
Climate maths 'doesn't add up' without carbon capture: COP28 chief
Climate maths 'doesn't add up' without carbon capture: COP28 chief
The Emirati oil chief leading this year's UN climate talks said Wednesday the world must get "serious" about new emission-capturing technology, rather than focussing only on replacing...
1970-01-01 08:00
GM’s Cruise to Expand Robotaxi Service to Dallas and Houston
GM’s Cruise to Expand Robotaxi Service to Dallas and Houston
Cruise, the self-driving vehicle unit majority owned by General Motors Co., plans to expand its robotaxi service to
1970-01-01 08:00
Bitcoin Drops Below $27,000 as Market Liquidity Concerns Climb
Bitcoin Drops Below $27,000 as Market Liquidity Concerns Climb
Bitcoin fell below $27,000 for the first time in more than two weeks, erasing an earlier gain posted
1970-01-01 08:00
The most popular search engine in the world is about to look different
The most popular search engine in the world is about to look different
Google is moving forward with plans to bring AI chat features to its core search engine as it works to keep pace with a wave of new artificial intelligence tools that could threaten the company's dominance online for the first time in decades.
1970-01-01 08:00
What Is Ray Tracing? (And What It Means for PC Gaming)
What Is Ray Tracing? (And What It Means for PC Gaming)
Noticed your eyes widening a lot more when you play games lately? If so, that
1970-01-01 08:00
A foldable phone, new tablet and lots of AI: What Google unveiled at its big developer event
A foldable phone, new tablet and lots of AI: What Google unveiled at its big developer event
Google on Wednesday unveiled its latest lineup of hardware products, including its first foldable phone and a new tablet, as well as plans to roll out new AI features to its search engine and productivity tools.
1970-01-01 08:00
Facebook Must Face Privacy Suit Over Cambridge Analytica
Facebook Must Face Privacy Suit Over Cambridge Analytica
A judge ruled Facebook founder Mark Zuckerberg and other former directors of the social-media company must face claims
1970-01-01 08:00
Binance’s Crypto Market Share Likely Fell After Zero-Fee Trading Halt
Binance’s Crypto Market Share Likely Fell After Zero-Fee Trading Halt
Binance appears to have lost market share amid changes to its trading-fee structure and a wider regulatory crackdown
1970-01-01 08:00
Hackers Contacted Cybersecurity Firm CEO’s Son, Wife in Extortion Attempt
Hackers Contacted Cybersecurity Firm CEO’s Son, Wife in Extortion Attempt
Hackers stole contracts from cybersecurity firm Dragos Inc. as part of an extortion attempt that included contacting the
1970-01-01 08:00
«453454455456»