Generating synthetic data could impose a 5x compute tax on model training
Dwarkesh Patel · AI Scaling, Alignment, and the Path to Superintelligence — With Dwarkesh Patel · May 15, 2024 · at 23:03
Dwarkesh Patel explains the technical compute overhead of generating and selecting synthetic data for AI model training to host Alex Kantrowitz.
0:00 / 0:18exact quote · 18.7s
▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark
“It will make training more expensive because instead of just doing one backward pass, you now potentially have to do many forward passes because at each forward pass, you're going to come up with some output. Then the model has to decide which of those outputs was the best. Now we're going to train on the best of those outputs. It could like Be a five X tax on training.”
quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →
More from Dwarkesh Patel
Patel: Individuals will eventually train superintelligences in basements
“The cost of training, the systems is declining so fast that Literally you will be able to train a super intelligence in a basement at some point in the future.”
Dwarkesh: AGI requires further algorithmic progress, not just current model scaling
“I don't think we're just right around our corner from AGI and it's just a little additional dash of something. That's all it's going to take. I think, you know, people often ask if all AI progress stopped right now and all you could do is collect more data or …”
Dwarkesh: Lack of continual learning prevents LLMs from replacing human labor
“I think a big bottleneck these models have is their inability to learn on the job, to have continual learning. Their entire memory is extinguished at the end of a session. There's a bunch of reasons why I think this actually makes it really hard to get human-l…”
Patel: Reinforcement learning may not generalize beyond verifiable domains
“I still think I I'm like, I'm not confident that this will generalize to domains that are not so verifiable or text-based.”