4 ms·
What would be my incentive to lie? I am getting absolutely breathtaking results from multi-task single shot prompts with one >1T model where I would need to fin
by treprinum 3y ago
What would be my incentive to lie? I am getting absolutely breathtaking results from multi-task single shot prompts with one >1T model where I would need to fine-tune dozens of models to get anywhere close, and likely with less reliability. While fine-tuning might be important to bring some benefits to smaller models, in the really big models CoT is often much better.
- chaxor 3y agoGeneral models are useful to the general public, and they're useful as a starting point to find tune to a smaller model. The purpose matters. Research and most products benefit heavily from having a much smaller model (speed) to do one thing very well on trillions of documents. General chatting from big models is mostly a gimmick, but they're still very useful to find tune.
- treprinum 3y agoThe thing is that those super large models with CoT are often much more reliable than specialized fine-tuned smaller models you'd like to use on trillions of documents, as I am observing right now. You are not trading just speed but also quality with smaller models. The only relevant use case for fine-tuning for me recently was to apply wizard dataset to LLaMA 2 for a less censored model and that was done using LoRA.
- byt143 3y agoWhat tasks?
- treprinum 3y agoFor processing trillion documents for example NER can be done much better.
- jerrygenser 3y agoYou are literally using trillion documents? Or are you exaggerating?
- treprinum 3y agochaxor above mentioned it so I quickly recalled a task I saw a super large LLM demolishing fine-tuned models on documents.
- byt143 3y agoIt's really depressing that a handful of big corporations will be able to exert such control over labor and productivity
- treprinum 3y agoThat's why the LLaMA 2 release is so significant. You can run the full 70B in 8-bit on two prosumer A6000 Ampere (cost around $10k together) which is within the reach of most companies and some devs. This could further accelerate all research to make it both efficient and available even to regular folks.
- amkkma 3y agoBut it's still not comparable to GPT 4, nor will it likely be for some time at least. And by the time we have GPT 4 class open source models, I'd imagine there'd be significant advancements in closed source models, such as inference time symbolic reasoning using MCTS, something google is working on for gemini...or just bigger/better architectures, data etc
- treprinum 3y agoYeah, I don't really see a solution. Even Stanford can't keep up with the latest AI research. Still, having LLaMA 2 is better than not having it.
- 3y ago
- fnordpiglet 3y agoI’d like to generate Minecraft fan fic. GPT4 does a terrible job because it’s been lobotomized. Fine tuning a model with a corpus of fan fic generates better fan fic. If I could fine tune GPT4 then ok I agree. But “general purpose lobotomized oracle” is only one use case for a language model, albeit a useful one, is the least creative one.
- treprinum 3y agoOK, understood. You might do that with uncensored versions of LLaMA 2 now. Role playing/fan fic is much better with those.
- fnordpiglet 3y agoThey tend to not have the right base of training materials - the models are constrained not just by the lobotomization but by some specialized writing not being part of the corpus. Training a lora on a specific corpus - fan fic, alt.sex.stories, or a persons specific writing, can help induce behavior or language choices specific to that corpus. You can also use context injection but I’ve found combining a model weight adjustment with priming the context to be the most effective. I think over time it’ll become standard practice to train special purpose models on top of general purpose models. If I have a very specific task domain a model tuned to that domain will less often wander off and will be more likely to respond in the desired way.