3 ms·
Just started listening to the Anthropic interview, and a few things jumped out at me. First, the claim that one of the main things holding LLMs back is a lack
by fumeux_fume 1y ago
Just started listening to the Anthropic interview, and a few things jumped out at me.
First, the claim that one of the main things holding LLMs back is a lack of expert feedback. To me, that just means the models are guessing—because they don’t have knowledge like humans do, they rely on pattern-matching, not understanding. If the user doesn’t know the answer, the LLM can’t help. That’s not just a minor limitation—it’s foundational. Framing it as a feedback issue is a way of sidestepping the deeper problem.
Second, the speculation about Claude winning a Pulitzer or a Nobel Prize. I get the underlying point—they're wondering whether LLMs are better at creative or scientific work. But couching it in terms of prestigious awards just adds to the hype. Why not just say “creative vs. scientific tasks”? Framing it as “what will it win first?” cheapens what those prizes represent and makes the model seem far more capable than it actually is.
Third, one of them claims a friend at a drug company says they’re about to release a drug discovered by AI. But when pressed for details, it turns out to be pure hearsay. It’s a textbook example of the kind of vague hype that surrounds LLMs—bold claims with no real substance when you dig.
That said, I appreciate that the host, despite being quite friendly with the guests, actually pushes back and holds them to their claims. That's not very common in AI discussions, and I respect it.
- terhechte 1y agoJust at note that > If the user doesn’t know the answer, the LLM can’t help. That’s not just a minor limitation—it’s foundational. doesn't mean the LLM is not useful. My favorite use case for LLMs is them doing something I know 100% how to solve myself. However, they do it much faster and I can 100% understand their solution and give feedback in case it is wrong. No surprises. I much prefer this as it allows me to work on multiple tasks in parallel.
- martingalex2 1y agoThis seems like an extreme of example of "all models are wrong, but some are useful".
- pfortuny 1y agoRight, you’ve just described a calculator.
- a_bonobo 1y agoYes!!! My favorite recent think pieces have been on LLMs as a 'normal' technology. Nothing revolutionary, no AGI, just a normal tool. An accountant knows how to do math but they use the math in Excel to do it for them. They 100% know how to do it themselves manually, but Excel does it far faster.
- arghwhat 1y ago> because they don’t have knowledge like humans do, they rely on pattern-matching, not understanding. I don't think this comparison makes sense - humans do not have knowledge as a unique thing differentiating us from LLMs. "Knowledge" is just the ability to produce relevant/"correct" results in some interaction, and some of us have acquired experiences that involve dealing with a particular subject and other knowledgable people within that subject for years or even decades, granting us (very fallible!) "knowledge" within that specific area. Humans hallucinate answers, even within their area of knowledge. Our memory is imperfect, so we're creating an answer based on bits and pieces, while anything missing is subconsciously glossed over and filled in with something plausible. It's just that the more experiences you have in an area, the better your hallucination. Those weird moments when your parents assure you of some fact that challenge the limits of just how wrong something can be would then be an example of hallucination in the extreme polar opposite end. (Note that I am not implying that the human brain works like an LLM, but rather just challenge the concept of "knowledge" being fundamenetally different from LLM behavior.)
- igouy 1y ago> Humans hallucinate answers Confabulate.
- arghwhat 1y agoThat does appear to be a better term for both humans and LLMs. Not sure who coined the term "hallucination" in the context of LLMs, which that suggests problems with perception rather than memory...
- computerthings 1y ago[dead]
- marginalia_nu 1y agoIf an AI correctly guesses the outcome of a dice roll, it does not mean it knows the outcome in advance. Being correct is a necessary element of knowledge, but far from sufficient. It's probably helpful to reach for epistemology to make sense of knowledge as a concept. While something like "true justified belief" is far from a perfect definition of knowledge, it a least is at least correct to a good approximation.
- roboboffin 1y agoHere is a link to the press release about the drug discovery: https://www.futurehouse.org/research-announcements/demonstrating-end-to-end-scientific-discovery-with-robin-a-multi-agent-system https://www.futurehouse.org/research-announcements/demonstra...
- whilenot-dev 1y agoPaper: https://arxiv.org/abs/2505.13400 https://arxiv.org/abs/2505.13400 Code: https://github.com/Future-House/robin https://github.com/Future-House/robin (NOTE: "will be available")
- seydor 1y agowe are at the peak of the LLM hype cycle. The trough of disillusionment will be a black bleak abyss
- ctoth 1y ago> First, the claim that one of the main things holding LLMs back is a lack of expert feedback. To me, that just means the models are guessing—because they don’t have knowledge like humans do, they rely on pattern-matching, not understanding. If the user doesn’t know the answer, the LLM can’t help. No. It means that the models don't have access to the physical world to run their own experiments. It's really hard to get smarter without feedback. Try it! You could have just posted this comment, then not gotten any pushback! And you would have been in the same pit of confabulation. But now feedback lets you correct your misapprehension. Cool init?