When I thought you were talking about a neural network that would take between a few GPUs and a few thousand GPUs to train:
This is what people are already doing, and the methods for discouraging dangerous outputs are not different from the methods used for discouraging other kinds of unwanted outputs. So just hire someone competent, who is keeping up with the state of the art, or make sure you are competent yourself. If your company is big, understand what the big companies do for safety; if your company is small, look into something like Meta’s Purple Llama.
When you clarified that you have a thousand times the compute used to train GPT-4, and your algorithm is much better than transformers:
That is not a recipe for an AI lawyer, that is a recipe for a god. If it works, it will take over you, then your company, then the world. For this, you don’t just need alignment, you need what OpenAI called “superalignment”. So maybe just sell your company to OpenAI, because at least they understand there’s a hard problem to solve here, that cannot be finessed.
The problem is that the risks involved with creating roughly human-level AI like GPT-4, and the risks involved with creating superintelligence, are quite different.
With human-level AI, we have some idea of what we are doing. With superintelligence, you’re a bit like a chimp breaking into a medical clinic. You might find something there that you can wield as a useful tool, but in general, you are surrounded by phenomena and possibilities that are completely beyond your comprehension, and you’ll easily end up doing the equivalent of poisoning yourself, injuring yourself, or setting the place on fire.
So I need another clarification. In the hypothetical, is your magic AI protocol capable of creating an intelligence much much greater than human, or should we only be concerned by the risks that could come from an entity with a human level of intelligence?
When I thought you were talking about a neural network that would take between a few GPUs and a few thousand GPUs to train:
This is what people are already doing, and the methods for discouraging dangerous outputs are not different from the methods used for discouraging other kinds of unwanted outputs. So just hire someone competent, who is keeping up with the state of the art, or make sure you are competent yourself. If your company is big, understand what the big companies do for safety; if your company is small, look into something like Meta’s Purple Llama.
When you clarified that you have a thousand times the compute used to train GPT-4, and your algorithm is much better than transformers:
That is not a recipe for an AI lawyer, that is a recipe for a god. If it works, it will take over you, then your company, then the world. For this, you don’t just need alignment, you need what OpenAI called “superalignment”. So maybe just sell your company to OpenAI, because at least they understand there’s a hard problem to solve here, that cannot be finessed.
Do you have a best guess though? Deferring is forbidden in the hypothetical.
The problem is that the risks involved with creating roughly human-level AI like GPT-4, and the risks involved with creating superintelligence, are quite different.
With human-level AI, we have some idea of what we are doing. With superintelligence, you’re a bit like a chimp breaking into a medical clinic. You might find something there that you can wield as a useful tool, but in general, you are surrounded by phenomena and possibilities that are completely beyond your comprehension, and you’ll easily end up doing the equivalent of poisoning yourself, injuring yourself, or setting the place on fire.
So I need another clarification. In the hypothetical, is your magic AI protocol capable of creating an intelligence much much greater than human, or should we only be concerned by the risks that could come from an entity with a human level of intelligence?