James W. A. Strachan, Dalila Albergo, Giulia Borghini
Oriana Pansardi, Eugenio Scaliti, Saurabh Gupta, Krati Saxena, Alessandro Rufo, Stefano Panzeri, Guido Manzi, Michael S. A. Graziano, Cristina Becchio
2024 · Nature Human Behaviour
Abstract At the core of what defines us as humans is the concept of theory of mind: the ability to track other people’s mental states. The recent development of large language models (LLMs) such as ChatGPT has led to intense debate about the possibility that these models exhibit behaviour that is indistinguishable from human behaviour in theory of mind tasks. Here we compare human and LLM performance on a comprehensive battery of measurements that aim to measure different theory of mind abilities, from understanding false beliefs to interpreting indirect requests and recognizing irony and faux pas. We tested two families of LLMs (GPT and LLaMA2) repeatedly against these measures and compared their performance with those from a sample of 1,907 human participants. Across the battery of theory of mind tests, we found that GPT-4 models performed at, or even sometimes above, human levels at identifying indirect requests, false beliefs and misdirection, but struggled with detecting faux pas. Faux pas, however, was the only test where LLaMA2 outperformed humans. Follow-up manipulations of the belief likelihood revealed that the superiority of LLaMA2 was illusory, possibly reflecting a bias towards attributing ignorance. By contrast, the poor performance of GPT originated from a hyperconservative approach towards committing to conclusions rather than from a genuine failure of inference. These findings not only demonstrate that LLMs exhibit behaviour that is consistent with the outputs of mentalistic inference in humans but also highlight the importance of systematic testing to ensure a non-superficial comparison between human and artificial intelligences.
Metadata: Crossref. Abstracts retain their original authors’ rights.
Original HN comments, with labels from simple text rules. A label describes the whole comment and may miss context or irony. Read the discussion to make up your own mind. About the labels
3 selected excerpts · 3 total mentions
But LLMs have been measured to have some theory of mind abilities at least as strong as humans: https://www.nature.com/articles/s41562-024-01882-z . At this point you either need to accept that either LLMs are already conscious, or that it's easy enough to fake being conscious that it's practically impossible to test for - philosophical zombies are possible. It doesn't seem to me that LLMs are conscious, so consciousness isn't really observable to others.
Needless to say, most humans are unoriginal parrots too, one need only look at the prevalence of memetic desire. Few are capable of artistic genius like Picasso. One technical definition of empathy is understanding what someone else is feeling. In war you must empathize with your enemy in order to understand their perspective and predict what they will do next. This cognitive empathy is basically theory of mind, which has been demonstrated in GPT4. https://www.nature.com/articles/s41562-024-01882-z If we do not assume biological substrate is special, then it's possible that AIs will one day have qualia and be able to fully empathize and experience the feelings of…
another. It could be possible that new AI architectures with continuously updating weights, memory modules, evolving value functions, and self-reflection, could one day produce truly original perspectives. It's still unknown if they will truly feel anything, but it's also technically unknowable if anyone else really experiences qualia, as described in the thought experiment of p-zombies.
>AGI would have to suffer to become general enough to warrant the use of the word general. Jo Cameron doesn't have general intelligence? This seems absurd. Intelligence is orthogonal to phenomenology and affective states. People aren't worried about AGI because it might have a Cartesian theater, the worry is that it might be more competent than humans and put humans out of a job. The semantics of whether it "truly" has intelligence is irrelevant. >Suffering is part of our intelligence and we’d be nowhere without it. It might be part of our intelligence, but why would it need to be part of machine's intelligence? GPT4 is already beating humans at theory of mind tasks[1], and I doubt that it suffers. Our suffering is an evolutionary stroke of bad luck. It has nothing to do with intelligence itself, and it would have been better if we had evolved some other way that didn't need it. [1] https://www.nature.com/articles/s41562-024-01882-z
More from the same topics.
Sébastien Bubeck, Varun Chandrasekaran, Ronen Eldan +11
Artificial intelligence (AI) researchers have been developing and refining large language models (LLMs) that exhibit remarkable capabilities across a variety of domains and tasks, challenging our understanding of learning and cognition.
Tom B. Brown, Benjamin Mann, Nick Ryder +28
Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task.
A. Pluchino. A. E. Biondo, A. Rapisarda
The largely dominant meritocratic paradigm of highly competitive Western cultures is rooted on the belief that success is due mainly, if not exclusively, to personal qualities such as talent, intelligence, skills, efforts or risk taking.
David Manheim, Scott Garrabrant
There are several distinct failure modes for overoptimization of systems on the basis of metrics.