Mozilla can't help it but be their own worst enemy. Ladybird may well never have happened if Mozilla just had kept working on Servo, and Ladybird is most definitely going to out compete Firefox when it reaches maturity, as Mozilla keeps on burning bridges with open source enthusiasts.
The problem with Mozilla is not just technical but cultural. The organization has been infected with managers. The managers want to keep their jobs more than they want Firefox to succeed. Clearly the solution is for the managers to fire themselves and allow the developers to run the show, but that was not going to happen.
Ladybird, by contrast, is a developer-lead open source project that has no such constraints. They also don't have a product yet but I'm sure the picture will be radically different in a few years.
> The problem with Mozilla is not just technical but cultural.
Not once in my career have I come across a problem that wasn't cultural. There are no purely technical problems in software. Everything can be achieved, everything can be worked around. All one need is a consensus. Enters cultural problems.
> The managers want to keep their jobs more than they want Firefox to succeed.
Coincidentally, also throughout my career, not once have I met an engineer that didn't put the entire blame on managers. Introspection really isn't our forte, is it? :)
I must have watched it at least 8 times, and only on the 9th time did I pause and realize that in this movie the hero and villain never meet. Willis and Oldman almost cross at the elevator but never actually meet.
This was a problem with early telephone lines which was easy to exploit (see Woz & Jobs Blue Box). It got solved by separating the voice and control pane via SS7. Maybe LLMs need this separation as well
This is where the old line of "LLMs are just next token predictors" actually factors in. I don't know how you get a next token predictor that user input can't break out of. The answer is for the implementer to try to split what they can, and run pre/post validation. But I highly doubt it will ever be 100%, its fundamental to the technology.
I think this is fundamental to any technology, including human brains.
Humans have a problem distinguishing "John from Microsoft" from somebody just claiming to be John from Microsoft. The reason why scamming humans is (relatively) hard is that each human is different. Discovering the perfect tactic to scam one human doesn't necessarily scale across all humans.
LLMs are the opposite; my Chat GPT is (almost) the same as your Chat GPT. It's the same model with the same system message, it's just the contexts that differ. This makes LLM jailbreaks a lot more scalable, and hence a lot more worthwhile to discover.
LLMs are also a lot more static. With people, we have the phenomenon of "banner blindness", which LLMs don't really experience.
So people can focus their attention to parts of content, specifically parts they find irrelevant or adversarial (like ads). LLMs on the other hand pay attention to everything or if they focus on something, it is hard to steer them away from irrelevant or adversarial parts.
Banner blindness is a phenomenon where humans build resistance to previously-effective ad formats, making them much less effective than they previously used to be.
You can find a "hook" to effectively manipulate people with advertising, but that hook gets less and less effective as it is exploited. LLMs don't have this property, except across training generations.
Maybe it's my failing but I can't imagine what that would look like.
Right now, you train an LLM by showing it lots of text, and tell it to come up with the best model for predicting the next word in any of that text, as accurately as possible across the corpus. Then you give it a chat template to make it predict what an AI assistant would say. Do some RLHF on top of that and you have Claude.
What would a model with multiple input layers look like? What is it training on, exactly?
It's hard in general, but for instruct/chat models in particular, which already assume a turn-based approach, could they not use a special token that switches control from LLM output to user input? The LLM architecture could be made so it's literally impossible for the model to even produce this token. In the example above, the LLM could then recognize this is not a legitimate user input, as it lacks the token. I'm probably overlooking something obvious.
Yes, and as you'd expect, this is how LLMs work today, in general, for control codes. But different elems use different control codes for different purposes, such as separating system prompt from user prompt.
But even if you tag inputs however your this is good, you can't force an LLM to it treat input type A as input type B, all you can do is try to weight against it! LLMs have no rules, only weights. Pre and post filters cam try to help, but they can't directly control the LLM text generation, they can only analyze and most inputs/output using their own heuristics.
Ah, I think this is due to human nature. People came in and wanted to "do" something with GitHub. To give it their stamp, to help boost their career/ego.
I bought a Zapos ZA-E200L in South-Africa for $65 [0]
It was configured with a fixed IP/subnet/gateway and no DHCP, which took an hour or two to debug and reconfigure. Been a dream since then: prints really fast and cuts properly. OP's project works well to print todo's.
This comes so timely. I bought the receipt printer last week and a large metal sheet to hold them magnetically to the wall for a real life todo kanban board
> Trump isn't doing anything out of the ordinary for an American president
I'm sorry, but I think both parties would actually agree on the fact that Trump is doing a lot of "out of the ordinary" for an American president.
No other president after WWII has reduced federal workforce by >8% (DOGE), and then rehired a bunch. No other US president ordered the capturing a head of state (Venezuela) and framed it as a law enforcement action. No president has ignored congress or the constitution like Trump has (tariffs, ICE, Greenland).
He uses executive orders a lot more than previous presidents: ~209 per year in his 2nd term. The next highest are Truman (113/year), Carter (80/year) and Kennedy (75/year).
Mozilla laid off the full Servo team, but never publicly announced this afaik. Wikipedia includes it here: https://en.wikipedia.org/wiki/Firefox#cite_ref-120
reply