To be honest, I tried to work with with GitHub Copilot to see if it could help junior devs focus in on the important parts of a PR and increase their efficiency and such, but... I've found it to be worse than useless at identifying where the real complexities and bug opportunities actually are.
When it does manage to identify the areas of a PR with the highest importance to review, other problematic parts of the PR will go effectively unreviewed, because the juniors are trusting that the AI tool was 100% correct in identifying the problematic spots and nothing else is concerning. This is partially a training issue, but these tools are being marketed as if you can trust them 100% and so new devs just... do.
In the worst cases I see, which happen a good 30% of the time at this point based on some rough napkin math, the AI directs junior engineers into time-wasting rabbit holes around things that are actually non-issues while actual issues with a PR go completely unnoticed. They spend ages writing defensive code and polyfills for things that the framework and its included polyfills already 100% cover. But if course, that code was usually AI generated too, and it's incomplete for the case they're trying to defend against anyway.
So IDK, I still think there's value in there somewhere, but extracting it has been an absolute nightmare for me.
I searched for the definition of "agent" and none of the results map to the way AI folks are using the word. It's really that simple, because we're marketing this stuff to non-tech people who already use words to mean things.
If we're redefining common words to market this stuff to non-tech people, and then we're conveniently not telling them that we redefined words, and thus allowing them to believe implicit falsehoods about the product that have serious consequences, we're being disingenuous.
That definition feels like it's playing on the verb, the idea of having "agency" in the world, and not on the noun, of being an "agent" for another party. The former is a philosophical category, while the latter has legal meaning and implication, and it feels somewhat disingenuous to continue to mix them up in this way.
Interesting. The best agents don't have agency, or at least don't use it.
You can think of this in video game terms: Players have agency. NPCs are "agencs", but don't have agency. But they're still not just objects in the game - they can move themselves and react to their environment.
That's actually a great example of what I'm saying, because I don't think the NPCs are agents at all in the traditional sense of "One that acts or has the power or authority to act on behalf of another." Where would the NPC derive its power and authority from? There is a human somewhere in the chain giving it 100% of its parameters, and that human is ultimately 100% responsible for the configuration of the NPC, which is why we don't blame the NPC in the game for behaving in a buggy way, we blame the devs. To say the NPC has agency puts some level of metaphysical responsibility about decision making and culpability on the thing that it doesn't have.
An AI "agent" is the same way, it is not culpable for its actions, the humans who set it up are, but we're leading people to believe that if the AI goes off script then the AI is somehow responsible for its own actions, which is simply not true. These are not autonomous beings, they're technology products.
In what way is it 'disingenuous'? You think Norvig is trying to deceive us about something? I'm not saying you have to agree with or like this definition but even if you think it's straight up wrong, 'disingenuous' feels utterly out of nowhere.
It's disingenuous in that it takes a word with a common understanding ("agent") and then conveniently redefines or re-etomologizes the word in an uncommon way that leads people to implicitly believe something about the product that isn't true.
Another great example of this trick is "essential" oils. We all know what the word "essential" means, but the companies selling the stuff use the word in the most uncommon way, to indicate the "essence" of something is in the oil, and then let the human brain fill in the gap and thus believe something that isn't true. It's techinically legal, but we have to agree that's not moral or ethical, right?
Maybe I'm wildly off base here, I have admittedly been wrong about a lot in my life up to this point. I just think the backlash that crops up when people realize what's going on (for example, the airline realizing that their chat bot does not in fact operate under the same rules as a human "agent," and that it's still a technology product) should lead companies to change their messaging and marketing, and the fact that they're just doubling down on the same misleading messaging over and over makes the whole charade feel disingenuous to me.
with a common understanding ("agent") and then conveniently redefines or re-etomologizes the word in an uncommon way that leads people to implicitly believe something about the product that isn't true.
Oh, I have no issue with his textbook definition, I'm saying that it's now being used to sell products by people who know their normal consumer base isn't using the same definition and it conveniently misleads them into believing things about the product that aren't true.
Knowing that your target market (non-tech folks) isn't using the same language as you, but persisting with that language because it creates convenient sales opportunities due to the misunderstandings, feels disingenuous to me.
An "agent" in common terms is just someone acting on behalf of another, but that someone still has autonomy and moral responsibility for their actions. Like for example the airline customer service representative situation. AI agents, when we pull back the curtains, get down to brass tacks, whatever turn of phrase you want to use, are still ultimately deterministic models. They have a lot more parameters, and their determinism is offset by many factors of pseudo-randomness, but given sufficient information we could still predict every single output. That system cannot be an agent in the common sense of the word, because humans are still dictating all of the possible actions and outcomes, and the machine doesn't actually have the autonomy required.
If you fail to keep your tech product from going off-script, you're responsible, because the model itself isn't a non-deterministic causal actor. A human CSR on the other hand is considered by law to have the power and responsibility associated with being a causal actor in the world, and so when they make up wild stuff about the terms of the agreement, you don't have to honor it for the customer, because there's culpability.
I'm drifting into philosophy at this point, which never goes well on HN, but this is ultimately how our legal system determines responsibility for actions, and AI doesn't meet those qualifications. If we ever want it to be culpable for its own actions, we'll have to change the legal framework we all operate under.
Edit: Causal, not casual... Whoops.
Also, I think I'm confusing the situation a bit by mixing the legal distinctions between agency and autonomy with the common understanding of being an "agent" and the philosophical concept of agency and culpability and how that relates to the US legal foundations.
Really hit the false equivalence nail on the head.
Cars enable long-distance travel, which in many parts of the world is now essential for survival, and often that is, ironically, a result of climate change.
LLMs do nothing to aid human survival. In a post-apocalyptic world, cars would be sought after for both transport and shelter, but nobody would give a care that we lost ChatGPT.
I took it as more of a recognition of the downsides to LLM tech that the hype train tends to ignore, as well as the legal / moral / ethical gray areas that exist in both training those models and determining who is liable for what they output.
> While they do this, they also use vast amounts of energy and are happy using all the GPUs you can throw at them which is a problem we’ve seen before in the field of cryptocurrencies.
How is that weird? Current energy consumption levels for LLMs are definitely problematic. That's not to say they can't improve, but at the moment it's definitely bad. The estimate of ongoing power consumption for Google's AI summary feature, if it was made available for every search result, is roughly equivalent to the consumption level of the entire country of Ireland. That's not awesome given most of the world is not powered by renewable energy sources.
Don't look at what it says, look at how it's written:
> While they do this, they also use vast amounts of energy and are happy using all the GPUs you can throw at them which is a problem we’ve seen before in the field of cryptocurrencies.
"Vast amounts of energy", "happy using all the GPUs you can throw at them", and then the random cryptocurrency reference to evoke all the bad stuff associated with those.
That's how a lot of folks write when they want to make a shorthand comparison and they trust their readers to understand what they're doing. The author is making a comparison of the current cost / benefit of LLM tech to the cost / benefit of Crypto. I don't necessarily agree with that comparison, but using an evocative writing style that can say quite a lot with very few words isn't a problem to me, especially given what this post is intended to convey.
Edit: I do think you're making a fair observation, I just feel there are a few reasons not to draw too strong of a conclusion from it
At a glance, it feels more likely to me that they're criticizing training LLMs on content without compensating the original creators, not the salaries of OpenAI engineers. It's a pretty common moral / ethical stance to take now.
That seems more likely than my initial interpretation, in which case the moral and ethical implications just so you can have a "Summarize with AI" button or other such features in your web browser are obviously much worse.
> It feels more likely to me that they're criticizing training LLMs on content without compensating the original creators,
No, the phrase they have written specifically talks about the RLHF workers and they do evaluate the pay of these people and somehow bring this issue into the discussion of how LLMs are useful to browsers.
And because I'm being downvoted for asking a simple logical question whether this has anything to do with LLM features in browser, I logically conclude that there are some things in current American culture that I literally (literally!) don't understand.
When it does manage to identify the areas of a PR with the highest importance to review, other problematic parts of the PR will go effectively unreviewed, because the juniors are trusting that the AI tool was 100% correct in identifying the problematic spots and nothing else is concerning. This is partially a training issue, but these tools are being marketed as if you can trust them 100% and so new devs just... do.
In the worst cases I see, which happen a good 30% of the time at this point based on some rough napkin math, the AI directs junior engineers into time-wasting rabbit holes around things that are actually non-issues while actual issues with a PR go completely unnoticed. They spend ages writing defensive code and polyfills for things that the framework and its included polyfills already 100% cover. But if course, that code was usually AI generated too, and it's incomplete for the case they're trying to defend against anyway.
So IDK, I still think there's value in there somewhere, but extracting it has been an absolute nightmare for me.