Mike Fraser

Sora getting shutdown means that Open AI was and still is struggling to find a use case for generalized AI applications. Meanwhile Anthropic and Google are zeroing in on practical applications for AI and taking market share.

This is mirroring the Internet bubble. Everyone rushed in with capital and over built a bunch of useless stuff. The bubble burst the cool stuff survived and everyone learned how to use the new tools properly.And then of course enshitified them.

Open AI is Pets.com

Mar 25 at 4:49pmWeb
Show threadMike Fraser4d ago
In the spirit of nostalgia for the dotcom era I should create an instance called worldcom.social
Show threadFeisty4d ago
@mike Oh, Pets.com…the ads were the best. https://m.youtube.com/watch?v=tMYvqRJiOd8
PETS.COM: Entire Campaign (Super Bowl ad) (Pets Dot Com)

YouTube
Show threadBrad Vrolijk 💬4d ago
@mike The real question will be whether the targeted applications will be enough for these companies to develop a profit model or not. From my friends in the dev industry, the extremes to which these tools need to be integrated/leveraged may not be realistic or sustainable.
Show threadMike Fraser4d ago
@buffaloseven I'm bullish on Anthropic as I see the demand coming from the bottom up. I think it'll be small business that will do the uptake and and the enterprise will follow but not necessarily lead. Dev teams right now are struggling because its a new paradigm just like Fortran programmers struggled when things moved off the mainframe.
Show threadBrad Vrolijk 💬4d ago
@mike Generally what I'm hearing is that context is a huge issue (which with enough dollars you can get around) and that broadly you can get function but not quality.
Show threadBrad Vrolijk 💬4d ago
@mike For context, though, I also believe that major gains with this iteration of LLM are done, and that ultimately the only broad skill improvement will be done by building bigger hammers, which works up to a point, but eventually is just diminishing returns.
Show threadcaffeinatedgeek 🇺🇦🇨🇦4d ago
@buffaloseven @mike I believe we will move from monolithic large models to a set of more specific models that an orchestrator engages. There will still be large models to address scenarios without a narrow model but that will become more rare.
Show threadKerry Stevenson4d ago
@mike I agree, but you do have to try things in order to see what works and what does not
Show threadMike Fraser4d ago
@krst Actually the bravery is in killing it. It's easy to keep dumping money into something that doesn't work hoping it will take.
Show threadKerry Stevenson4d ago
@mike esp throwing away $1B from the Disney deal in the process
Show threadBowreality 🇨🇦🍁4d ago
@mike Have they learned nothing ?!
Show threadAmin Girasol4d ago
@mike pets.com with added necropolitics.