zlacker

Doesn't the existence of consumer products like ChatGPT indicate that LLMs aren't able to do human-level work? If OpenAI really had a digital workforce with the capabilities of ~100k programmers/scientists/writers/lawyers/doctors etc, wouldn't the most profitable move be to utilize those "workers" directly, rather that renting out their skills piecemeal?

>>caditi+(OP)
That depends on what the real value is. The sure way to get rich selling pickaxes to gold miners. However you would be even richer if you figured out where the gold really was and mined in that exact location.'

Of course you can also get rich selling scams.

>>caditi+(OP)
If this stuff worked, then given that Microsoft has a huge share of it, shouldn't Microsoft's products be good? Or at least getting better. I use Bing most days [1], and it's consistently an absolute joke.

[1] to farm reward points to get cosmetic items in video games

>>caditi+(OP)
LLMs are able to do human-level work. I think surely now that is unquestionable? They can't do all human-level work, in the same way that not all humans can do all human-level work, but a lot of criticism implicitly benchmarks it against all work. LLMs have this remarkable ability to be both the smartest thing in the room and the dumbest thing at the same time.

On the profitability point. Yes and no I guess. ChatGPT will likely go down as the most influential thing in the ai race (or early ai race) simply because it was the most visceral illustration of the art of the possible to the general public.

Technical people will correct me or have arguments to why this isn't the case but they'll be wrong simply because ChatGPT captured consumer attention, which ultimately led to supercharged financial attention, and everything that has happened since is due to that. Whether its pure marketing or futuregazing, the world of money bought into it, with the ultimate view that this is the worst it will ever be.

The ultimate profitable goal is unarguably replacing human labour (at a lesser all in cost) but if that is not possible, but you believe it could be possible, then showcasing and getting people to believe that you can achieve that goal, is pretty damn profitable too and it's proven to be for some people while giving them runway to continue to pursue the ultimate goal.

>>cal_de+QY
> ChatGPT will likely go down as

It's been a long time since I thought of the Segway, which exemplifies one possible trajectory: Recognizable to all, everyone interested in trying one out, but still fell short of the investor dream of replacing the car and changing urban travel forever.

>>Terr_+B21
Sure but my point isn't that ChatGPT will be dominant, it has already changed the world. Even if Openai fell over tomorrow, that doesnt change that it was the step change from a public consciousness in the potential of LLMs. And LLMs aren't falling by the wayside anytime soon

>>caditi+(OP)
For now it's better and easier to do this with our money and our feedback.

But yes, someone else said that as soon as any of these get to AGI you will not be able to use it.

>>bluGil+M2
All the truer when there is more miners than gold, one can't help but note.

>>glemio+kl1
To be more exact, the fact that everyone sells access to models they are continually "improving" is proof that any statement from the LLM peddlers about AGI is a lie.

>>nottor+if2
Why?

It funds the initial cost and reduces risks

It also provides additional training data and feedback

A run can cost 100 Million $

>>cal_de+QY
What we really need is an updated Turing test: "can a project manager tell the difference between a junior dev working remotely vs Claude Code hooked up to a Slack account?"

>>caditi+KP4
Or "can a junior dev tell from an email or Jira ticket whether it was written by Claude Code hooked up to email/Jira or an actual PM" =)