List of All Articles with Tag 'h'

No sooner did ChatGPT get unleashed than hackers started “jailbreaking” the artificial intelligence chatbot – trying to override its safeguards so it could blurt out something unhinged or obscene

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

By Timothy Gardner WASHINGTON (Reuters) -Private U.S. nuclear fusion company Helion Energy will provide Microsoft with electricity in about five

Much of Europe and the northeast US are in store for sweltering summer temperatures well above historical norms,

When the Ethereum blockchain moved away from using a technique for verifying transactions known as proof of work

Space is full of mysteries which we will never likely solve but, generally, we at least know a thing or two about craft we send out into the cosmos. Not so, when it comes to one probe which has just touched back down on Earth after 276 days in orbit. The experimental spacecraft was launched by China’s space agency, ostensibly to test the nation’s reusable space technologies. According to state media agency Xinhua News, the mission’s aim was to help with the development of “more convenient and affordable round-trip methods for the peaceful use of space in the future”. Sign up for our free Indy100 weekly newsletter That all sounds pretty similar to what SpaceX and NASA are up to, except that the Chinese spacecraft and its journey have been shrouded in secrecy. No information has been released on the altitude it reached or the systems it tested. We don’t even know where it went or, indeed, what kind of spacecraft it was – not a single image has been released to the public. Commentators on Chinese social media have speculated that Beijing has been developing a spacecraft like the UAir Force's X-37B, an autonomous spaceplane that can remain in orbit for years. However, no one knows how well this is going nor, indeed, if it’s going at all. All we do know about China’s latest unidentified flying object is that it was launched from the Jiquan Satellite Launch Center in the Gobi Desert on August 5, 2022 and returned to the same site on 8 May, 2023. It follows an earlier mission, carried out in July last year, which saw a Chinese spacecraft fly to the edge of the Earth’s atmosphere and back on the same day. The country’s main space contractor, China Aerospace Science and Technology Corp (CASC), hailed the success of the craft’s brief celestial jaunt at the time. It gushed: "The development of reusable space transportation technology is an important symbol of China's transition from a 'big' space-faring nation to a 'powerful' space-faring one.” Have your say in our news democracy. Click the upvote icon at the top of the page to help raise this article through the indy100 rankings.

Concerning messages showing on people’s phones have led to fears that they are being listened to. The app appears to be attempting to access the microphone within people’s phones, when users are asleep, according to the privacy menus within Google’s Android operating system. But WhatsApp said that it believes the issue is a bug within Android, and has stated categorically that the app would not access user’s microphones without their permission. The problem was raised by Twitter engineer Foad Dabiri, who shared a tweet which included a screenshot of a page in Android that shows when a given app accesses the microphone. He noted that it showed a range of attempts to do so, even when he had been asleep. The post was further amplified by Mr Dabiri’s boss, Elon Musk, who shared the tweet and claimed that “WhatsApp cannot be trusted”. Mr Musk’s post followed a range of other criticisms of Mark Zuckerberg, largely for political reasons, and Twitter also competes with WhatsApp with its direct messaging platform. In another post, Mr Musk pointed to the fact that WhatsApp is “owned by Meta/Facebook”, and claimed that WhatsApp’s former owners had left the parent company “in disgust”. “What they learned about Facebook and changes to WhatsApp obviously disturbed them greatly,” he claimed. Other users also reported seeing the same problem, on Twitter and Reddit, even before Mr Dabiri brought it to widespread attention. The screen can be seen within Google’s “Privacy Dashboard”, which is found within the Settings app and collects information about how personal data is being used by apps on a phone. Some claimed that page even showed WhatsApp accessing their camera. Others said that the microphone appeared to be getting accessed every few minutes. The same privacy dashboard can be used to turn off the microphone for any apps that users may be concerned about. But that could lead to certain features not working, such as voice notes or calls on WhatsApp. WhatsApp said that it believed the screen was showing the microphone being accessed as a result of a “bug”. It said that it was showing the wrong information in Google’s privacy dashboard screen and that it had been in touch with Google to fix it. “Users have full control over their mic settings,” WhatsApp posted in response to Mr Dabiri’s tweet. “Once granted permission, WhatsApp only accesses the mic when a user is making a call or recording a voice note or video – and even then, these communications are protected by end-to-end encryption so WhatsApp cannot hear them.” Meta-owned apps have long been plagued by fears that they are accessing the microphones of users. For years, users have accused Facebook and Instagram of listening to them to collect data and show them more relevant sponsored ads, which Meta has always firmly denied. Many experts have argued that the sometimes uncannily accurate sponsored posts are simply the result of the app’s advertising tools accurately profiling users, without needing to listen to them. Read More WhatsApp just fixed two of its most glaring quirks Google to unveil major new AI Quantum computer discovers bizarre particle that remembers its past Google to unveil major new AI Quantum computer discovers bizarre particle that remembers its past Why is Elon Musk purging Twitter accounts?

Guerrilla Games is celebrating the impressive sales of 'Horizon Forbidden West'.

A series of revelations in recent weeks about Supreme Court Justice Clarence Thomas accepting lavish vacations and private-school

By Akash Sriram Shares of Rivian Automotive Inc rose 7% on Wednesday as its positive earnings stood out

By Jeffrey Dastin Alphabet Inc's Google on Wednesday is expected to unveil more artificial intelligence in its products

Overwatch 2 fans can get their hands on some free rewards this May through special Twitch drops. Here's what's up for grabs, and when you can earn them.