No real disagreement on the fact of apps. It's just that they don't need VC money for a chance to discover product-market fit--the barriers to entry are low-enough that you can bring products to market fairly quickly. The flipside is that the apps are unlikely to have enduring, capital-lite pricing power . . . because the barriers to entry are too low. Bootstrapping will be the new-normal.
re: AI uses cases, can't recall if I posted this here before, but there is a page talking about using AI to fix journalism to make it more efficient to allow more local news outlets (i.e. transcriptions of public meetings and summaries, other automated filtering, which to the public may be better than no local news at all), as well as AI to assist mainstream media to regain the public's trust. Niche media can be biased, but mainstream claiming to be neutral would benefit from having AI help nudge journalists to be more neutral, e.g. explaining to a progressive journalist how a conservative would react to their article. In the real of software there is the concept of "pair programming" where two programmers work together: this would be "pair journalism" except with a human paired with an AI. Its here:
Though its unclear the potential revenue, which is true for many AI applications at the moment, especially "apps" are merely thin layers over the big models which risk lots of competition from many chasing the same low hanging fruit that comes to mind. Tech in search of a problem. The issue is whether there are problems that have been searching for answers that AI will can solve and where there is some way to grab the niche early and develop a competitive advantage.
Re: Andreessen’s comment, he doesn’t seem to be making a reasonable argument. He immediately assumes the conclusion and asserts it forcefully, and then proceeds to psychoanalyze his opponents and denounce their alleged motives. I’m surprised you find this kind of discourse acceptable, let alone agreeing with it.
Does Andreessen believe that AIs will never reach a superhuman level of intelligence, or is he confident that they’ll remain under human control even if they do? In his rush to denounce his opponents, he’s not even pausing to make it clear which position he’s arguing for. (Unfortunately, I’m afraid that if he did, he would hardly be able to make a convincing case for either.)
Great update. I used the Teach-A-Thon concept for the first Chat GPT training in my company. It was a good framework. I also agree on the apps on top of LLMs. The key differentiator though is going to be data the LLMs don’t have. You can use their framework but need to bring unique data to the table or your app has the gather it fast enough to give you some kind of moat.
No real disagreement on the fact of apps. It's just that they don't need VC money for a chance to discover product-market fit--the barriers to entry are low-enough that you can bring products to market fairly quickly. The flipside is that the apps are unlikely to have enduring, capital-lite pricing power . . . because the barriers to entry are too low. Bootstrapping will be the new-normal.
re: AI uses cases, can't recall if I posted this here before, but there is a page talking about using AI to fix journalism to make it more efficient to allow more local news outlets (i.e. transcriptions of public meetings and summaries, other automated filtering, which to the public may be better than no local news at all), as well as AI to assist mainstream media to regain the public's trust. Niche media can be biased, but mainstream claiming to be neutral would benefit from having AI help nudge journalists to be more neutral, e.g. explaining to a progressive journalist how a conservative would react to their article. In the real of software there is the concept of "pair programming" where two programmers work together: this would be "pair journalism" except with a human paired with an AI. Its here:
https://FixJournalism.com
Though its unclear the potential revenue, which is true for many AI applications at the moment, especially "apps" are merely thin layers over the big models which risk lots of competition from many chasing the same low hanging fruit that comes to mind. Tech in search of a problem. The issue is whether there are problems that have been searching for answers that AI will can solve and where there is some way to grab the niche early and develop a competitive advantage.
Re: Andreessen’s comment, he doesn’t seem to be making a reasonable argument. He immediately assumes the conclusion and asserts it forcefully, and then proceeds to psychoanalyze his opponents and denounce their alleged motives. I’m surprised you find this kind of discourse acceptable, let alone agreeing with it.
Does Andreessen believe that AIs will never reach a superhuman level of intelligence, or is he confident that they’ll remain under human control even if they do? In his rush to denounce his opponents, he’s not even pausing to make it clear which position he’s arguing for. (Unfortunately, I’m afraid that if he did, he would hardly be able to make a convincing case for either.)
Great update. I used the Teach-A-Thon concept for the first Chat GPT training in my company. It was a good framework. I also agree on the apps on top of LLMs. The key differentiator though is going to be data the LLMs don’t have. You can use their framework but need to bring unique data to the table or your app has the gather it fast enough to give you some kind of moat.