I believe if he tried, he actually could make denying the cookies an even worse experience.
Anyway, to the point, sure, AI is overblown way out of proportions. But it’s useful and is here to stay. Like, does someone actually believe that a software that can clone a voice in seconds is going away? Or a software that can generate an image of whatever you want in seconds? Or a software that can predict how valuable your customer is going to be in a year and thus predicting which customers are actually worth spending money on?
Is AI a bubble? Yes. But anyone who thinks that it’s comparable to the crypto bubble doesn’t understand AI, or crypto, or (most likely) both.
Not one of his better articles. A lot of prevaricating with very little substance or conclusions. While AI investment is most certainly a bubble, few are taking into account the technology improving. This is a bone the tech dogs aren’t going to give up easily. They have a whiff of the possibility of AIG becoming a reality. Like all promising tech, the potential for long term damage is as, if not more so, catastrophic for society as the World Wide Web and social media. He’s looking at this as an echo of past Silicone Valley bubbles. It so much more than that.
Very little substance or conclusions. While technology is improving, you’re not reading into account AI investment is a bubble.
AI can certainly help, but not a single one was able to consistently deliver good results. A technology that needs constant supervision by an actual expert isn’t really all that useful. And this is not just a problem of scale. It’s a limitation of the current approach. Throwing billions at a problem to save a few millions just isn’t worth it.
They have a whiff of the possibility of AIG becoming a reality.
They also had a wiff of NFTs letting them sell and claim royalties on JPEGs. This isn’t about some grand vision if humanity’s future, it’s about becoming the next Silicon Valley billionaire, or dethroning the richest man in the world. If the next big tech get-rich-quick scheme comes along, the novelty of their very expensive autocomplete and JPEG mashup projects will be dead, and they’ll take their dollrs on to the next fad.
He’s looking at this as an echo of past Silicone Valley bubbles. It so much more than that.
It’s really hard to know how this will play out. The models only have to improve a bit at this point to be reliably better than humans, as which time it probably makes sense to replace humans. It seems they will probably still hallucinate but do it little enough that it’s still a net gain to use them. Compute power needed to run them will surely come down.
I’m as skeptical as the next guy, but I do think they will have uses, especially in examples like radiology which he he uses as a negative case. However I’m pretty sure it will eventually be able to do the initial screening to find the 95% of cases with nothing at a rate similar to existing medical diagnostic testing and then return the other 5% back to a human to review and decide further treatment. Based on my experience with speech language models, I’m pretty sure you’d be able to tweak the models to produce mostly false positives rather than false negatives and then run it through further layers of review afterwards.
the electricity bill for each query – to power the servers and their chillers – would still make running these giant models very expensive.
This assumes there won’t be radical advances in cost-effective hardware to run the queries.
AI proponents work precisely towards such advances: hardware tailored to running the best performing models, at far lower costs than current GPUs and GPU derivates.
Something like “execute in RAM” neural network accelerators, could reduce query costs by several orders of magnitude.
The fraud of the cryptocurrency bubble was far more pervasive than the fraud in the dotcom bubble, so much so that without the fraud, there’s almost nothing left.
Ironically, what the crypto bubble left behind, was a surplus of GPUs, which got repurposed for AI… and just like crypto left GPUs to move onto purpose-built ASICs and other models like PoS instead of PoW, so does AI need to leave GPUs and move onto purpose-built hardware with better models (quantized NNs are a good example).
As a tangent, we could talk about how the gaming industry has enabled a GPU industry that in turn has enabled these off-shots.
I didn’t wanna read the article nor the opinions in it, so I asked ChatGPT.
AI can be seen as a bubble in the sense that there’s a lot of hype and excitement surrounding it, often leading to inflated expectations. However, there’s also substantial substance and potential for real impact. It’s important to navigate through the hype to understand both its capabilities and limitations.
Damn imagine paying to read an article to confirm your biases. If you’re going to claim something is a bubble you need to claim something more specific than “AI”.
The entire early modern web was a bubble in the early 00s and it’s still here. There’s not even many large companies yet to even start being bubbly. An actual AI bubble could be 10 years off.
I just love how so many educated and intelligent people can get stuck in knee jerk reactionary takes because some admittedly large aspects of a new technology (LLMs) are annoying as hell.
This bubble is quite bubbly. There is an AI company for anything and everything now. The market is almost fully saturated with “AI” everything.
Just like the web bubble, all of the intsta-AI shops need to fail so the real tech can grow. AI is never going to go away, but most of the scam companies will fail in due time.
We might have one big consolidation, or several. The hype will die and the quick money will disappear. It’s the same story, every time. One the magic AI box stops shitting out dollar bills, we should be good.
It surely is a bubble - so probably a bit different than many other bubbles.
I think OpenAI made the right call (for them) to commercialize when they did - as that pretty much was their only chance to do so. Things has moved fast over the last 1.5 years - and what used to take a decade in tech has happened within months: OpenAI is the dinosaur company grandfathered in, while for already about a year it’s been more sensible for anybody wanting to do something with LLM to selfhost (or buy hosting capacity, but put up own data) one of the more open language models, and possibly adjust or re-train it.
As a company owner I get a ridiculous amount of spam for a year already from all kinds of companies building products on top of OpenAI stack, or are trying to sell training or conferences. All those companies will be left with nothing once all the slower users realize technology has moved on. It’s like somebody trying to build all their product offerings based on VMWare stack nowadays.
If you as a company want to offer something around AI right now the safest option is probably offering hosting, or if you want to do more hands on, adjustment of open models. Both of those are very risky, and many will go bust in years to come - but not as suicidal as building on top of a closed dinosaur.
A point the article makes rather well is that something is not a bobble because it doesn’t work, but because the investment going into it is fundamentally irrational in scale. The web still existing has nothing to do if investment or companies tripping over themselves to advertise as a dotcom in the dotcom bobble was rational, percicly because it clearly wasn’t dispite the web being a fundamentally revolutionary tech.
The question when it comes to LLM’s, the near exclusive subject of the marketing around AI, is if bunch of random companies paying for a mildly improved chat bot are actually going to generate enough profit once the marketing hype has worn off and the legal challenges settled to justify the current massive scale of investment, or if instead once the project managers and CEO’s have moved on to the next buzzword to attract investors LLM’s will become a tight market where providers struggle to turn a large enough profit to satisfy investors.
So far, the only thing AI has shown to be pretty good at is summerizing a large amount of data, and even then it cant be fully trusted to not make mistakes.
AI is a tool, and one need to learn using it. But even today, just two years from introduction to the public (LLMs) it has lot's of uses. This is one of the fastest adoption of nearly any technology.
Hmm, I think summarization is a bad example. I’ve read quite some AI summaries that miss the point, sum up to a point where the simplification makes sth wrong or the AI added things or paraphrased and made things at least ambiguous. Even with the state of the art tech. Especially if the original texts were condensed or written by professionals. Like scientific papers or good news articles…
What I think works better are tasks like translating text. That works really well. Sometimes things like rewording text. Or the style-transfer the image generators can do. That’s impressive. Restoring old photos, coloring them or editing something in/out. I also like the creativity they provide me with. They can come up with ideas, flesh out my ideas.
I think AI is an useful tool for tasks like that. But not so much for summarization or handling factual information. I don’t see a reason why further research coudn’t improve on that… But at the current state it’s just the wrong choice of tools.
And sure, it doesn’t help that people hype AI and throw it at everything.
I’d say it isn’t, but I’d be saying it about the existing applications (and near-future stuff that currently sorta-kinda works), not the tech-bro delusions that wind up on billboards and cause aging suits to both sweat and drool.
AI is the kind of bubble where the tech involved is halfway to magic. And it’ll run on your local hardware. No matter how hard the offices-and-capital side crashes and burns, that’s not going anywhere.
Right now is the worst that image AI will ever be again.
LLMs might stumble, because the big-iron approach seems to make a difference for them, but there are local versions and they do roughly the same things. That’s going into video games, for a start, and probably turning every single NPC into a verbal chatbot that can almost hold a conversation.
The universe of low-stakes, high-dollar applications for AI is so small that I can’t think of anything that belongs in it.
… entertainment is a bajillion-dollar industry where abject failure is routine and hilarious. There will be a boom like there was for CGI, only with even worse treatment of the tiny companies performing miracles. Workers getting screwed while revenue floods in is not the same thing as a bubble bursting. Unfortunately.
I’m disappointed in Doctorow for asserting this technology will remain big and complex and expensive. When has that ever stayed true? Saying it’ll always take forest-eating big iron sounds like predicting computers will only be affordable to the five richest kings of Europe. This whole neural-network revival kicked off because consumer hardware made training feasible. More training equals better models, and more computers equals more training, but Google’s still pouring gigawatts into glorified video-game tech.
If all the creative and academic zeal gets left working with mundane single machines - guess where all the advancements will happen.
Didn't NASA also announce something like this in the last couple of months? I couldn't find anything online but I swear I saw something about NASA promoting a similar initiative recently.
These are some pretty abysmal winners. Kaiju Preservation society is at best a pulpy beach read. Haven't read Babel but I was unimpressed by Poppy Wars and the excerpts I've seen from Babel lack coherent voice.
locusmag.com
Hot