LLMs Get Better One Task at a Time Now, Not All at Once
- Source
- x.com
- Date
As amazing as LLMs are, improving their knowledge today involves a more piecemeal process than is widely appreciated. I’ve written before about how AI is amazing... but not that amazing. Well, it is also true that LLMs are general... but not that general. We shouldn’t buy into the inaccurate hype that LLMs are a path to AGI in just a few years, but we also shouldn’t buy into the opposite, also inaccurate hype that they are only demoware. Instead, I find it helpful to have a more precise understanding of the current path to building more intelligent models. First, LLMs are indeed a more general form of intelligence than earlier generations of technology. This is why a single LLM can be applied to a wide range of tasks. The first wave of LLM technology accomplished this by training on the public web, which contains a lot of information about a wide range of topics. This made their knowledge far more general than earlier algorithms that were trained to carry out a single task such as predicting housing prices or playing a single game like chess or Go. However, they’re far less general than human abilities. For instance, after pretraining on the entire content of the public web, an…
Context
A 's broad range of uses can conceal how much targeted work goes into improving particular abilities. In Ng's account of model development, Andrew Ng describes researchers collecting or generating domain-specific examples, cleaning them and preparing them for training.
For abilities such as operating a browser, Ng describes constructing simulated environments where a model can repeatedly practice a narrow set of tasks. Ng's argument is that progress still depends heavily on human choices about data and practice conditions, and that models remain far less general than people even after training on most of the public web. Ng expects this incremental work, even if breakthroughs do not change it, to keep driving rapid progress.
- LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
- pretraining — The first, biggest phase of building a model: training it on enormous amounts of text so it learns language, facts, and reasoning in general.
Checking sign-in…
Loading comments…