Any script kiddie with an LLM could write a worm that would spread through the supply chain, possibly even hacking speakers right on the factory floor and blasting Rickroll music or something similar.

It would be interesting to see if Creative would still claim that it "does not present a cybersecurity risk".

Edit: Bonus points for closing the security hole and disabling the ability to flash the firmware normally, so that the manufacturer would have to jailbreak the speakers in order to repair them.

Edit: I haven't gone through all the code, but they might do something like this: https://arxiv.org/abs/2211.17192 where a draft model is used and the output distribution is tweaked on rejection, resulting in the exact same distribution as the main model.

As far as I know, this is not used in practice. Currently popular implementations always match the main model output, and the draft model only affects the speed.

The draft model quickly generates draft-token 1.

The main model then starts working on two tokens in parallel. It calculates token 1 based on the context, and token 2 based on the context + draft-token 1.

Once the two tokens have been generated, you can check whether the draft-token 1 from the draft model matches token 1 from the main model.

If they match, you have just calculated two tokens in the time it takes to generate one, because the calculation was done in parallel. If they do not match, delete token 2 and generate it again. Since you have already generated the correct token 1 with the big model, you can use the context + token 1 (from the main model). This takes more time, but the result is always the same.

The draft model essentially predicts the next token quickly, enabling you to start generating the subsequent token in parallel. If the guess is right, the second generated token is correct. If it is wrong, the second generated token is also potentially wrong, so it must be generated again using the correct prior token obtained through the big model.

A poor draft model will simply slow down the process without affecting the output.

If you decide to swerve, the additional weight at the front will help you to initiate the turn, and good systems will then reduce the braking force at the right moment to give you the most traction when cornering.

It corrects spelling errors and improves awkward wording. You can then go and choose alternative sentences or words. Just don't expect any sort of deeper intelligence.

If an end-to-end model is used and there is no second, more traditional safety self-driving stack, like the one Mercedes will use in their upcoming Level 2++ driving assistant, then the model can be manipulated essentially without limit. Even a more traditional stack can be vulnerable if not carefully designed. It is realistic to imagine that one printed page stuck on a lamppost could cause the car to reliably crash.