TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

AI isn’t good enough

172 点作者 MaysonL超过 1 年前

36 条评论

ambrozk超过 1 年前
This entire piece is based on one massive, unsupported assertion, which is that LLM progress will cease. Or, as the author puts it, "we are at the tail end of the first wave of large language model-based AI... [it] ends somewhere in the next year or two with the kinds of limits people are running up against." I want to know only one thing, which is what gives him the confidence necessary to say that. If that one statement is untrue, the thesis of the piece completely fails, and I do not know how any one person alive today can be certain that that statement is true. Has Paul Kedrovsky or Eric Norlin spent $150M training a 2T parameter model that no one's heard about? Do they have access to classified OpenAI docs which state that GPT5 exists already, and it's no better than GPT4? Without this sort of information, "LLMs are not going to get smart enough for widespread practical use" is an unsubstantiated bet, and nothing more.
评论 #37259285 未加载
评论 #37259556 未加载
评论 #37263356 未加载
评论 #37258752 未加载
评论 #37258382 未加载
评论 #37258921 未加载
评论 #37258404 未加载
评论 #37258253 未加载
评论 #37262708 未加载
评论 #37260792 未加载
jillesvangurp超过 1 年前
By the time AIs stop hallucinating, it will be effectively super human (i.e. much better) because humans hallucinate all the time. And some people are barely coherent. We&#x27;re holding AIs to a much higher standard than ourselves. And we move the goal posts all the time as well.<p>Let&#x27;s deconstruct that title. &quot;AI isn&#x27;t good enough&quot;. Good enough for what? Great example of moving the goal posts. Because anytime it fails to do whatever, it&#x27;s not good enough. But what about all the stuff it is good enough for that people are actually using it for already? It&#x27;s passing tests at a level most humans only manage very briefly after lots of intensive preparation. The knowledge slips away quickly after that.<p>The way I see it, there&#x27;s a long, rapidly growing list of stuff that AIs have nailed already and a list of things where it is clearly struggling. That list is shrinking.<p>Self driving gets cited a lot as something where AIs is failing. It&#x27;s hard to say that in a world where multiple companies are now operating self driving vehicles commercially in several cities across the US, China, and soon Europe. Are they perfect? No. Do they drive better than the average jerk on the road? Definitely; not even close. Most people simply suck at driving. That&#x27;s why traffic deaths are so common. Most of those deaths are caused by somebody with a drivers license proving that they shouldn&#x27;t have received one. Not quite a solved problem but at this point the question of when these things will be on the road in my city is more a matter of when than if. It already works in other cities. I don&#x27;t see what&#x27;s so special in mine that it couldn&#x27;t be replicated here. In other words, I expect the number of cities with autonomous vehicles to start exploding over the next few years. And who knows, Tesla might hit the good enough mark at some point as well. Again, the barrier is very low because humans aren&#x27;t held to the same standards. They&#x27;ll give drivers licenses to just about anyone.
评论 #37260177 未加载
dontupvoteme超过 1 年前
I am completely bewildered by the responses here.<p>GPT-3 is so good it could write code, or it could emulate an entire TV show to the point of lawsuit, or simulate an eternal debate between Herzog and Zizek. You could translate Bulgarian to Sindarin for heaven&#x27;s sake. And all of this comes at a 50% tax rate or so because of the alignment tax that OAI opted for. (which doesn&#x27;t even work because anyone who puts effort in can jailbreak it scarily well)<p>The real solution is not these monstrosities of the internet mushed together and taxed like a Belgian billionaire. It&#x27;s models and frameworks specific to what you want in a given query.<p>It was less than 24 hours ago that we finally got a local model that can do code generation well, and we also know that Meta has a far better one that they are holding back.<p>There is no reason that for any query you should be restricted to a single inference run on a single fixed model, and there is no reason that we shouldn&#x27;t perform a bunch of non-LLM processing on the output before the user sees it.<p>Switch between multiple models, fine-tune, or use LoRAs based on the query -- Python plots? Load llama-coding-python-plots. You like Plotly? Add -Plotly. Run the code in a sandbox (with hard kills on resources) and regenerate it if it does not meet standards.<p>You&#x27;re working in linguistics? Switch to a chat model that&#x27;s fine-tuned in the literature and codebase of that field.<p>There are a trillion ways to improve things; we&#x27;re basically at the cavemen-banging-rocks-together point.<p>Hell, I&#x27;m a sleep deprived ESL and i just invoked the monstrosity to fix my grammar and spelling for this post. What a bloody waste of CO2.<p>If for nothing but to keep the Earth afloat we shouldn&#x27;t be using these huge closed source SaaS things unless we need them. Repurpose them for SETI@Home or processing human genetics for healthcare, or something else that benefits mankind
atleastoptimal超过 1 年前
&quot;we are already at the tail end of the current wave of AI.&quot;<p>Is that really true? Generative AI spending this year just so far is 4x greater than last year, and the results of that spending haven&#x27;t been released yet. How can they conclude this?<p>Most &quot;predictions&quot; about the future I&#x27;ve seen assume that ChatGPT will stay the same and not improve for the next 10 years, or only marginally. Why they make this assumption, I don&#x27;t know, it seems to go against every past trend we&#x27;ve seen with machine learning scaling. There are even people who say LLMs have reached their limit because we&#x27;ve exhausted all the text data that exists, assuming that the only way for LLMs to improve is to blindly feed it more static data and that no other sources of improvements exist.
评论 #37258343 未加载
评论 #37259986 未加载
评论 #37259553 未加载
ironborn123超过 1 年前
Rather than asserting that current LLMs are at their tail end, or that AI isnt good enough, it is much more instructive to check what are the bottlenecks or constraints to further progress, and what would help remove these bottlenecks.<p>They can largely be divided into 3 buckets<p>1) Compute constraint - Currently large companies using expensive nvidia chips do most of the heavylifting of training good models. Although chips will improve over time, and competition like Intel&#x2F;AMD will bring down prices, this is a slow process. But what could be a faster breakthrough is training using distributed computing over millions of consumer GPUs. There are already efforts in that direction (eg. petals&#x2F;swarm parallelism for finetuning&#x2F;full training, but the eastern europe&#x2F;russian guys developing them dont seem to have enough resources).<p>2) Data constraint - If you just rely on human generated text data, you will soon exhaust this resource (maybe GPT4 has already). But the Tinystories dataset generated from GPT4 shows if we can have SOTA models generate more data (and especially on niche topics that appear less frequently in human generated data), and have deterministic&#x2F;AI filters to segregate the good and bad quality data thus generated, data quantity would not be an issue any longer. Also, multimodal data is expected (with the right model architectures) to be more efficient at training world grokking SOTA models than single modal data and here we have massive amounts of online video data to tap into.<p>3) Architectural knowledge constraint - This may be the most difficult of all, figuring out what is the next big scalable architecture after Transformers. Either we keep trying newer ideas (like the stanford hazy research group does), and hope something sticks, or we get SOTA models few years down the line to do this ideation part for us.
jongjong超过 1 年前
As impressive as AI is and what it can do. I&#x27;ve felt the same way in my last job. It was almost there; it was almost good enough such that we could almost ship the project but we didn&#x27;t. GPT3.5 was not good enough, GPT4 seemed like it was good enough in terms of accuracy but it was too slow and we were getting throttled like crazy. I lost my job over it. When I was worried about AI taking my job, I didn&#x27;t picture it like this.
评论 #37257897 未加载
评论 #37261294 未加载
评论 #37262064 未加载
ryanklee超过 1 年前
&quot;We are quickly reaching the limits of current AI&quot;<p>So many words in this piece, but nothing to back up this primary assertion.
评论 #37258389 未加载
评论 #37258448 未加载
评论 #37257799 未加载
评论 #37258597 未加载
nmca超过 1 年前
This is nuanced and interesting, despite most likely being completely wrong - as others have mentioned, it&#x27;s predicted on LLMs being at the top of the sigmoid and they just ain&#x27;t.<p>Also, this goal-post-moving did crack me up rather:<p>&gt; Sure, it is terrific at using its statistical models to come up with textual passages that read better than the average human’s writing, but that’s not a particularly high hurdle.
countWSS超过 1 年前
Data-centric token-guessing will be outperformed by more complex architectures, LLMs are too inefficient and their hardware requirements are disproportionate for the utility. Building a sand castle of &quot;LLM apis&quot; that rely on plugging their deficiences is a fundamentally bad decision that hides the underlying waste of resources. What will be great if the input data it was fed was more structured and labeled like images: imaging a &quot;Book, 17th century, reliablity:54%, contains geographic errors&quot;. Instead this &quot;attention is all you need&quot; approach treats garbage data equal to top material.
borissk超过 1 年前
Very interesting article. IMHO a key factor in how the world changes in the next 10-20 years will be how far the current wave of AI, based on neural networks can go.<p>If scientists and engineers manage to implement autonomous cars and robots that can do basic human tasks like clean, wait and take care of the sick and elderly we will wake up in a brand new world. It will put a lot of stress on the basics of society, such as democracy and having to work for a salary.
评论 #37258109 未加载
评论 #37257991 未加载
评论 #37258546 未加载
评论 #37258325 未加载
Nemi超过 1 年前
I hope beyond hope that we continue to iterate on the current GPT-4 model. But statements made by Altman himself make me leery that it might not happen. That, coupled with the fact that it is just so expensive to run make me wonder if this whole thing is sustainable.<p>Take this into consideration - NVDA’s massive profit in the last couple of quarters is someone else’s expense. When no one is making money on AI, can this continue? I absolutely think companies can make money <i>using</i> AI that people will pay for, but I am not convinced that companies can make money <i>supplying</i> AI services. I see it as almost a commodity type service, like internet access or something. There is no clear path to profitability, and that is ok to some extent, but with the massive costs this could come crashing to a halt very quickly.
YeBanKo超过 1 年前
With few prompts I just wrote a small script with a bunch of openssl commands. Saved me an hour. And it works. And it even explained it. There is AGI and then there is “practically agi”. The latter has been achieved with LLMs.
arisAlexis超过 1 年前
&gt;We are bumping against many of its &gt;limits<p>He keeps saying that as a fact with absolutely no proof and based his whole argument on it.
评论 #37261683 未加载
评论 #37261316 未加载
michaelmrose超过 1 年前
&gt; But almost everywhere else needs people. Badly. Across retail, restaurants, manufacturing, trades, and on and on, companies are struggling to hire. And this brings us back to the cross-country trip observation and that “signing bonus” sign in a McDonald’s window.<p>For jobs that pay just enough to not qualify for food stamps or Medicaid but not enough to buy food medicine and rent.<p>Previously it was more possible to earn little enough to qualify for benefits and live if not well but the numbers have become misaligned and don&#x27;t account for differences between expensive and cheaper parts of states.<p>Given present concerns the situation doesn&#x27;t change appreciably unless you go all in on automation implement UBI or both.
petermcneeley超过 1 年前
&quot;too few people for all the jobs, for the most part&quot; - Are wages going up?<p>&quot;But almost everywhere else needs people. Badly. Across retail, restaurants, manufacturing, trades, and on and on, companies are struggling to hire. &quot; - Wow sounds like an exciting star trek future that I cant wait to be part of!<p>&quot;labor became more expensive than capital&quot; - So wages must be going up right?<p>&quot;In essence, the authors show that for automation to have widespread benefits it must deliver high productivity gains to more than compensate for the high displacement of workers.&quot; - the reason why this sentence makes no sense is because it is written in two frames at the same time. One of a planned economy and the other of capitalism.<p>&quot;We are in a middle zone, however, with AI able to displace huge numbers of workers quickly, but not provide compensatory and broader productivity benefits.&quot; - But what about the horses?<p>I tried my best to find the central thesis of this rambling multi-tone work. I think it is difficult because it was written with too many different frames.
评论 #37257693 未加载
评论 #37259023 未加载
评论 #37258222 未加载
评论 #37257909 未加载
评论 #37259484 未加载
评论 #37257980 未加载
评论 #37258119 未加载
xpe超过 1 年前
I’m so done with writing that has an implicit but uncommunicated quantitative model. (AI costs are too high? Relative to what? Show us!) Almost all analytical &#x2F; business &#x2F; industry writing should quantified.<p>I hope for better. We all should. It is within reach. Yes, I&#x27;m grumpy on this topic, but this doesn&#x27;t make it unwarranted; it is a <i>consequence</i> of wanting better for a decade or longer.<p>At this point, if you claim to be a business analyst and you aren&#x27;t sharing a simulation, logic, or code, I&#x27;m just not going to take you seriously.
intrasight超过 1 年前
&gt; too few people for all the jobs, for the most part.<p>Change that qualifier to &quot;for a minority part&quot; and it would be correct.<p>There are plenty of workers - old and young and everything in between - who would gladly work if a) you allowed them to and&#x2F;or b) you paid them enough.
Madmallard超过 1 年前
GPT4 has been beyond our expectations by a mile. What in the world is this person even on?
tlogan超过 1 年前
I do not think it is fair to use AI in support call centers as an example.<p>These call centers are not there to help customers (cancel subscription, refund, change something, etc.) and AI is perfectly tuned for that. Or at least they have procedures and AI is perfectly train to follow these.<p>In other words, AI in support centers is designed and works to follow companies procedures and rules—which may not always align with customer expectations.
评论 #37262507 未加载
manuhortet超过 1 年前
Great article. Although I am not sure if we really are reaching the value limits of the current LLMs, as the author mentions. Some fields as legal litigation are being disrupted as much as call centers have, but the change is happening in a much slower fashion, for many reasons. The productivity gain in these fields is potentially high, but so is human displacement. Relevant fact to consider
wildermuthn超过 1 年前
The most interesting part of this article isn’t its assertion that AI innovation is about to slow, but rather that current LLMs can only produce “so-so automation” — that although LLMs can replace many jobs, LLMs aren’t good enough to offset the cost of people losing such jobs.<p>True as of today, but the article is overlooking multi-modal LLMs. Once AI can see, we’ll go from “wow, GPT4 is pretty good at helping me to my job” to “uh oh, GPT4-Vision can do my job better than I can”.<p>This is especially true if the model can read text off a screen. At that point, it’s likely to replace many workers who stare at screens for a few hours a day and pretend to work the remainder. The model won’t take coffee breaks, or long lunches, or take the afternoon off to drive their kid to the dentist. If you’re a worker whose work depends upon looking at a screen, and mostly involves moving data from one place to another with light transformations and interpretations, you’re going to be in trouble by the end of the decade.
评论 #37262731 未加载
abdussamit超过 1 年前
I haven&#x27;t read the article (like many here) and I don&#x27;t get why the writer would write this. LLMs&#x2F;AI are not a stage where they can replace people or their jobs (broadly speaking), they still have to be taught and engineered to solve your problems. For example, if I enter a prompt &quot;Write a Javascript function that does XYZ for me.&quot;, and don&#x27;t get a desired result, it&#x27;s unfair for me to say &quot;ChatGPT bad, AI bad&quot;.<p>At this moment, you have to work with the limits, use better prompts, use ChatGPT as a guide and not as your personal robot. With that approach, I feel lots of revolutionary content is incoming in products. Use the tool better to get the most out of it.
philipwhiuk超过 1 年前
We don&#x27;t need better LLMs to eliminate all McDonalds workers. In fact there&#x27;s no language model needed there at all. The job just needs better robotics.
评论 #37263797 未加载
Ekaros超过 1 年前
How far are we from a point you could leave AI to prompt something itself. That is not just respond to something, but actively seek it.<p>Let&#x27;s say use this place. Could it without being ordered read articles, comments and add it own comments? Maybe search internet by itself for relevant content and post it? Maybe try to earn maximal amount of upvotes? And all this without there being any bit of code poking it to do it?
评论 #37259992 未加载
评论 #37260028 未加载
fullstackchris超过 1 年前
AI won&#x27;t work to reeingineer society, just like the last 100 &quot;latest things&quot; didnt. There is a massive undercurrent that always goes in the opposite direction of technology advancement: tradition!<p>You need new people growing up and living in a new paradigm to get these changes, you cant just ask (or expect) people who have lived their life one way to change completely to some other way of living
paulpauper超过 1 年前
<i>The absence of human workers has become a limiting factor on economic growth in the U.S.</i><p>the problem is real GDP growth is flat or negative due to high inflation. having more labor would make this worse, if anything. job creation tends to be inflationary. from 2010-2021 saw strong economic growth and low inflation, but now things suddenly changed due to too much inflation, not a lack of growth.
评论 #37258540 未加载
jy1超过 1 年前
People often conflate AI progress with hardware progress.<p>If we had hardware as flexible, robust, and &quot;cheap&quot; as the human body, we&#x27;d solve a lot of these problems such as retail, trades etc.<p>The problem isn&#x27;t software. It&#x27;s hardware.
seydor超过 1 年前
Just hook the LLMs to a robot arm and watch another, much larger wave of AI innovation
ericksoa超过 1 年前
When I see VCs make this argument, all I hear is &quot;oh shit, we&#x27;re late to the party&quot; and doing something like this as a desperate attempt to bring valuations down.
w_t_payne超过 1 年前
We&#x27;re just not that good at <i>managing</i> AI yet.<p>Current agent-based systems are little more than toys ... but this situation is improving very rapidly.
Havoc超过 1 年前
I don&#x27;t think the assertion that we&#x27;ve reached the end of current wave of AI is correct. It hasn&#x27;t even been rolled out to <i>anything</i> yet.<p>Once it is fully integrated into os&#x2F;browser&#x2F;corporate workspace and the MBA&#x27;s have thrown LLMs at every business problem they can think of then perhaps but right now it hasn&#x27;t even left high tech circles
评论 #37259625 未加载
zoogeny超过 1 年前
When I read articles like this I can&#x27;t help but question my own intuitions since they are so opposed to this author&#x27;s thinking.<p>Firstly, his suggestion on the worker shortage and the anecdotal $1000 sign on bonus at some local McDonalds. These seem more tailored to a narrative interpretation of the state of the world that is formed to the author&#x27;s own bias. I look at my own small town homeless and drug addicted population and consider the changes in the last 5-10 years on that front. I consider the insane rise in the cost of home ownership and the massive rise in rental costs locally. I just don&#x27;t see $1000 signing bonuses at fast food restaurants as a sign of some latent demand for labor. It feels like the author is using a single anecdote and a narrow reading of a few graphs to support a world view that I don&#x27;t see existing in my own experience.<p>Then he says: &quot;We are quickly reaching the limits of current AI&quot;. You mean the AI that surpassed most informed peoples expectations on AI capability <i>less than one year ago</i>? I feel like I&#x27;m in 1996 reading an article about the Internet where someone gives some list of reasons why the Internet is reaching its limits.<p>And just as an aside, this author uses the term &quot;explosive&quot; especially in the context of &quot;explosive growth&quot; over and over. It may be that his target market just isn&#x27;t me. The narrative and analysis probably hits home for a particular set of capitalist minded investor class individuals. Perhaps they really do see North America as having a huge labor shortage and that we just need the &quot;right&quot; kinds of automation so we don&#x27;t all feel so bad about the continuing displacement of workers. As long as they keep seeing &quot;explosive growth&quot; in the value of their assets.
speak_plainly超过 1 年前
“Historically, when this has happened—labor became more expensive than capital—economies have responded with automation, so we should expect that again today.”<p>Foolish assertion. Immigration has always been the stop-gap solution for expensive labor to the point that it’s built in as an essential feature of liberalism.
rldjbpin超过 1 年前
disclaimer: i am in the middle of the journey into learning this subject, and have at least done finetuning on existing models. some may still consider my take as misinformed.<p>one thing i am not a big fan of in this discourse is that how the community makes up jargon to describe relatively simple things on one hand (a la &quot;embeddings&quot;), and while come up with misnomers like &quot;hallucinations&quot; to describe otherwise blunt things. it is contributing to miss the forest for the trees in this discussion.<p>while i agree it is quite impressive to see what the state of the art can seemingly achieve on surface, the way we approach designing these models are fundamentally not aligned with what we want to do with its result. while some media for generative machine learning produce relatively harmless results (e.g. images and audio), it is the text where i have a lot of reservations.<p>we train these models to essentially turn noise into something that humans can pass as acceptable (for the given prompt). Sadly, however, people outside (and sadly &quot;inside&quot;) this space claim it to be the word of God.<p>i still strongly believe that a hybrid system is going to be the only way we can achieve the most from the current approach. almost a decade ago we were already good at parsing through sentence structure and &quot;understanding&quot; what the user writes. what the output is can easily be done with some hardcoding in logic with some noise thrown in by the models. however, it should only be treated with that level of seriousness.<p>i might be shouting into a void, but it is imperitive that we stop taking what these models output as production-ready code, or as a starting point even! otherwise, the only thing you achieve from this is to train the models running the chat services further.
asu_thomas超过 1 年前
&gt; <i>There is a persistent structural imbalance in the U.S. workforce: too few people for all the jobs</i><p>Lies. If the minimum wage increased proportional to inflation, none of these &quot;jobs&quot; would exist.<p>The McDonalds sign-on bonus is nothing but a loan, just like all sign-on bonuses with a retention requirement. Such arrangements merely require that minimum wage is low enough to partition that compensation out of the base wage, a modern American financial industry to efficiently enforce the contracts, and a society with no remaining semblance of dignity in sight.
version_five超过 1 年前
Labor shortage it BS. It&#x27;s rhetoric to increase immigration so we can continue sustaining a slave class and don&#x27;t have to pay people properly. Canada&#x27;s been very successful at this.