Reasons:
- OpenAI lobbying trying to get open source competition banned
- Apple releasing the M4 with those hardware specs
- Big brands like Adobe finding early success
- Industries are ripe for disruption: Medical, legal, procurement, communications of various kinds especially around documentation, support, summarizing, education, media, art, music, film and other entertainment, I can imagine a lot of disruption.
What would this do to Netflix:
render a 100-hour Harry Potter film that follows every detail of the books exactly
Or: render a Tom Cruise sci-fi movie starring Sigourney Weaver in her prime and Bruce Willis as the villain
How could Netflix or even YouTube keep up with scripts that render any video you can imagine? They'll have to offer it too, or go the way of Blockbuster. I imagine record labels and film studios will sue and try to get certain models banned that imitate celebrities or film styles, but people will just illicitly share models with torrents. There's going to be a whole era, I think, because the tech will need to be captured by the big boys but it's not there yet - so part of that is letting the market come up with ideas before they absorb it all then burst the economy.Our search engines and websites will have LLMs imbued into their functionality, smartphones will have them, kiosks IRL will have them, SaaS will use them to replace representatives and online order forms.
IMO AI models are potentially more disruptive than social media and crypto because it's not just the sharing of content unrestricted, it's the unrestricted creation of content and information itself - giving basically anyone the ability to do it.
What do we really need LLMs for that humans can't?
I would argue that it's not about whether or not we need something that humans can already do, it's that we don't want a human to do those things because a machine might be able to do it better, faster, cheaper, etc.A human can operate a lathe, I used to do it, but the CNC lathe I also operated could do any thing I could do much, much faster with just as much accuracy.
This year
> What will trigger it?
World events affecting all markets
> What do we really need LLMs for that humans can't?
LLMs are fantastic and are a huge boon. We'll be using them to our collective advantage for a long time to come. But the AI bubble is more a reflection of capital movement than technological promise.
Remember Cisco. Still not above the peak.
No, everyone will be in exactly the same doghouse. Nvidia has experience hedging their bets on multiple silicon suppliers, I suspect they are better-prepared for this situation than any other manufacturer.
Maybe Apple will try to rekindle that partership with ARM, or maybe they've diversified enough that they can try to make something with old TSMC fabs in Vietnam, assuming the war doesn't engulf them too, but I didn't think any sub 10nm stuff was being made outside of Taiwan.
The invasion is more or less a certainty in the next few years, though. I am much more sure about that than any prognostications about Apple's response to the invasion
[0] https://arstechnica.com/tech-policy/2023/07/tsmc-delays-us-c...
The bubble might burst though when there are big break throughs by someone like OpenAI or google, as right now money is being thrown at anything with AI in the name, when those startups fail to deliver the bubble will burst and all the spoils will go to a few that actually are able to deliver, making the rest irrelevant.
- consumers using chat & image gen
- devtools buying, and users leveraging, tools like kapa.ai
- RAG definitely provides ROI
- medical imaging & drug discovery
- robotics like Stanford Aloha
- ecosystem for day-2 problems in LLM based systems
The applications are only going to grow
There was also a real estate bubble in the early 00s with their 0-money-down loans, infinite refis, the "good faith estimate" - they were throwing money (houses) at people because middlemen could make a lot of money off selling loans. We all know about the burst, but few talk about the fact that the big banks took over those businesses, forcing them to shut down and acquired all their assets - they scooped up what worked and left everything else behind.
It seems like every presidential cycle there's some industry or space getting all the focus and funding. This current one - Trump/Biden - was the weirdest one, we had a brief crypto bubble, but it's been mostly Old Money paying themselves through government and politics (covid, wars, "keep the government running by printing $1T" type bills). I imagine the next president will "save us" from the political bubble and embrace technology, in particular AI/LLMs. The major corporations have to get a handle on AI, so I predict a lot of money will be thrown at AI/LLMs - we're just getting started. I think we have 8 years of AI bubble ahead of us.