it appears to be a version of gpt-4o that’s been trained on answers that explicitly show their logic. openai is claiming the results can’t be match by prompt engineering alone
Misunderstood or marketed in a dishonest way? The stakes are incredibly high for OpenAI and I think you make a good point that OpenAI also doesn’t really know what o1 is or isn’t good for. It’s presented as a leap, but is it? I think THEY don’t even know but surely hope so — because it has to be.
Marketed by OpenAI, misunderstood by the rest of us. Anecdotally it seems to be pretty good at medical or legal advice (hallucinations notwithstanding), in a clear departure from 4o
Fun toy for researchers, but hard to justify $2400 a year for the neglible amount of personal efficiency gained that for most people appears to be partially based on belief rather than evidence. Can't see many employers forking out for their employers either, especially since Apple has ads that show the AI powered employee as a lazy ass who gets paid to tell a machine to do all their work.
If o1's capabilities are meaningfully useful to someone, the $200/month is absolutely worth it. The main issue is that the vast majority of ChatGPT users, to your point, are going to see negligible benefits from o1 vs GPT-4o. It's not meaningfully better at writing emails or summarizing call transcripts.
But I think that's generally the case for high-end tools: if I'm biking to the gym and back, there's no point in shelling out for a $50,000 bicycle. If I'm trying to compete in professional races, however, it's a different story.
Has the underlying method for determining responses changed in the model or is it just showing how it got to the response , like showing your working?
it appears to be a version of gpt-4o that’s been trained on answers that explicitly show their logic. openai is claiming the results can’t be match by prompt engineering alone
Okay interesting, why would it be different to prompt engineering results?
Misunderstood or marketed in a dishonest way? The stakes are incredibly high for OpenAI and I think you make a good point that OpenAI also doesn’t really know what o1 is or isn’t good for. It’s presented as a leap, but is it? I think THEY don’t even know but surely hope so — because it has to be.
Marketed by OpenAI, misunderstood by the rest of us. Anecdotally it seems to be pretty good at medical or legal advice (hallucinations notwithstanding), in a clear departure from 4o
My reading is a bit more skeptical but we’ll see how it holds up against real use cases. In the end, the proof is in the pudding ;-)
Fun toy for researchers, but hard to justify $2400 a year for the neglible amount of personal efficiency gained that for most people appears to be partially based on belief rather than evidence. Can't see many employers forking out for their employers either, especially since Apple has ads that show the AI powered employee as a lazy ass who gets paid to tell a machine to do all their work.
If o1's capabilities are meaningfully useful to someone, the $200/month is absolutely worth it. The main issue is that the vast majority of ChatGPT users, to your point, are going to see negligible benefits from o1 vs GPT-4o. It's not meaningfully better at writing emails or summarizing call transcripts.
But I think that's generally the case for high-end tools: if I'm biking to the gym and back, there's no point in shelling out for a $50,000 bicycle. If I'm trying to compete in professional races, however, it's a different story.