So if I'm reading this correctly, leftmost seems justified to finetune your own LM, but middle seems at risk for next GPT to outperform the task-specific model?
@_jasonwei
-
Examples of Fine-tuned Models in Recent Research
By
–
Hm, well I mentioned PubMedGPT and GPT-3 finetuned on GMS8K in the post, does that count? If so, I'm sure I could find plenty more examples by skimming recent arxiv
-
Foundation Model Training: When Fine-tuning Isn’t the Right Choice
By
–
I guess a case here to not finetune is if you are training your own foundation model and you can't easily finetune the next SOTA base model
-
The Future Role of Task-Specific and Finetuned Models
By
–
An open question these days is what the role of task-specific / finetuned models will be. I can only think of three scenarios where it makes sense to work on task-specific models. The first scenario is if you have private (e.g., legal, medical, business) data not found on the
-
First Waymo Experience: SF Feels Futuristic with Self-Driving Taxis
By
–
Lost my waymo virginity last night; it was the first time I felt like SF is a futuristic city (though also dystopian, with all the homelessness around). I don’t think I’ll take human taxis again if I have the choice. Self-driving taxis drive more smoothly and are cleaner (and
-
Benefits of Pair Programming in AI Research and Development
By
–
Pair programming isn’t standard at most companies and basically non-existent in academia, but I’ve been doing it with @hwchung27 for almost a year now. While it naively seems slower to code individually, I’ve realized that there are many benefits: (1) In AI, what you work on can
-
Top AI Leaders Balance Management with Individual Contributor Technical Work
By
–
It seems to be not a coincidence that some of the strongest leaders in AI who manage large teams frequently do very low-level technical work. Jeff Dean doing weekly IC (individual contributor) work while managing 3k+ people at Google Research is the canonical example, but I've
-
Definition of Emergent Abilities in Large Language Models
By
–
My personal opinion is that if you explicitly train for something but it only works for large-enough models, then I'd consider that emergent. Here is an example from https://
arxiv.org/abs/2112.00114 which is cited in our emergent abilities paper. -
Complete Transparency with Your Manager: A Game-Changing Policy
By
–
One thing that I started doing at OpenAI is that I created a policy for myself to be *100% transparent* with my manager about everything. It seems obvious and weird to say aloud, but I bet most people don’t actually do this. But once I started doing it, I realized there are a lot
-
Girlfriend as Reward Model, Boyfriend as Policy Model in RLHF
By
–
My girlfriend doesn’t like the weekend plans I make for us, but she also doesn’t want to make plans herself. Instead, I should propose multiple schedules and then she picks one she likes. So I said she is like a Reward Model in RLHF, and I am like a Policy Model (with a low LR).