A big problem is these political spectrums are not really coherent moral philosophies yet fulfill an individual’s need for that kind of structure. I think anyone would find that “a person in America has a larger carbon footprint than a person almost anywhere else; therefore it’s an ecological imperative to expel whomever we can“ is odious. So we just don’t think about it.
Yes, and here towards the bottom the author gets to the reason he didn’t sign with the other medalists:
> I felt that there was nothing to be gained from criticizing AI companies for generating too many solutions too quickly.
> Under the circumstances, I think the best we can do is recognise the changes that are coming and try to work out the least unsatisfactory way of dealing with them.
Basically let’s make it a short-term problem and deal with it. Groups of people can deal with short term emergencies. Don’t turn it into a structural issue.
And in my view what’s the alternative in the letter exactly? The tools exist. Is there going to be drama every time somebody decides to use them?
The tools exist, so let’s try to figure out as humans the best way to use them to promote humanity, and let’s advocate against ways of using them which are a detriment to humanity, which is exactly the charge the letter makes. There are cultural norms around the use of every technology.
I'd like to hear how your comment isn't irrelevant noise, then. How does one implement these grand ideas except by state action? Otherwise, people will do what people will do, including proving theorems by machine.
Well I think in response to 'what are you going to do', what people are probably going to do is stop sharing on-going work, stop bothering to curate problems that are just going to inflate someone's IPO valuation, basically accelerating the tragedy of the commons that the AI companies are driving.
Seems legit. I speculate their are emergent mechanisms for the development of memory that are not dependent on what ever mechanism is behind the "improve $model for 'everyone'"*
Is someone knowlegeable on research that engages the question on the development of something that could be considered as some emergent mechanism of memory, that is independent of instances and their contextwindow an specifically also independent of the use of conversations for trainingsdata?
*(This is regarding the value of selfhosted models - maybe small selforganisations that share ressources to do though to do so? Generally the value of machine learning seems to be to big to reject)
Their privacy policy for normie subscribers says in plain English they use your Personal Data for research. I think it’s pretty unreasonable to use the service and expect otherwise.
My doctor's privacy policy is a bit more abusive than OpenAI's. It exists mostly because of a $%^&&* legal framework rather than malice, and I've grown accustomed to "if I don't want to die then I sign away these rights." Despite my having theoretically signed my soul away, my doctor isn't selling personal information to my exes or to life insurance companies (though they could in the US; that extremely personal information is no longer mine). OpenAI is engaging in the "technically legal maybe we'll see but obviously unintended" side of this transaction, and maybe that works out for them, but I wouldn't personally choose to be a shill for "it's unreasonble to expect somebody with 'legal' permission to do something other than the maximum 'legally' permitted" if I were in your shoes.
I have the setting turned on in Gemini Pro even though I work on proprietary code because 1. the setting allows for some (very limited) "memory", and 2. I consider my source code almost public even when it's not open source because I don't work on programs that involve extremely high level of know how or proprietary algorithms. It's mostly CRUD that can be copied in a myriad of ways, whether people use my methods or other methods.
If Gemini can improve based on my code and sessions (maybe doubtful but who knows) and others can benefit from it, that would be a welcome side-effect.
Depends on your profession? I sometimes work on open source code as part of my professional duties. Nothing that goes on there is necessary to keep private.
Aren't business consulting firms even worse? They are explicitly for business and there are known cases where they shared confidential information of one of their customers with another one.
> You'd think theft would still be illegal regardless of what a privacy policy says.
I wish that were true, but I live in the United States and it is 2026.
The President of the United States rug-pulls memecoin crypto and regularly pardons people like Paul Walczak (who was convicted of massive payroll fraud) in exchange for large donations.
I wouldn't make any assumptions about what is considered theft anymore, at least not when it is being committed by people who have enough money to be above the law.
You could probably train this out. I don’t think you need to develop elaborate filters. It doesn’t seem like that big a hill to climb if it’s important to people.
That's why this paper is important - it shows it isn't trained out. Leaving no other information in the model makes it clear what the biases are, and that the model is willing to make a biased decision. If you give it other unbiased criteria as well the bias may still easily remain but not be as clear.
Not sure it’s that strong. The prompt gives the presumption that this matters. Not necessarily a training issue vs the prompts being poorly written and the results being inherent in the bias they carry
Eh, I tried in 2017 with BA when their system went down and received nothing. There was magically a Black Lives Matter protest at Gatwick (or maybe LCY, can’t remember) on the same day, which isn’t covered apparently.
Almost certainly not. The legislation says: "An operating air carrier shall not be obliged to pay compensation [...], if it can prove that the cancellation is caused by extraordinary circumstances which could not have been avoided even if all reasonable measures had been taken."
That gets them off the hook for even second order effects. No harm throwing in a claim though, there's no penalty for having a claim turned down.
Isn’t their whole thing supposed to be spying on foreigners? They seem to be quite successful. There aren’t that many exchanges [1]. Could probably manage with cash, guns, and some know-how.
If you just look at the largest 4 of those, you'd have 100Tbps of traffic to monitor, with an average throughput of roughly half of that.
That's ~540PB ((50 Tbps / 8 bits) * 86400 seconds/day) of traffic a day with just those four. Add in the rest and you're likely talking ~Exabytes of data each day. And that has to all be processed on site.
If someone wants to argue that the NSA is in these facilities I'd be 100% onboard. But inspecting it all would be nearly impossible, let alone capturing it all and sending it back to some datacenter somewhere, which is a physical impossibility.
That's nothing a rack full of fast switches can't handle. A rack full of fast switches already does handle it - where do you think the original copy came from?
They will get a copy of the whole feed, but not store all of it - they will have heuristics for selecting interesting traffic.
Switches handle data at far faster rate than any hardware can actually inspect it, store it, process it, etc.
But yeah, just a rack of "fast switches" is all it takes to route hundreds of petabytes of data each day. You should let the data center operators know. They'd save billions.
Switches do inspect it. They also have a feature designed for wiretapping, which copies a percentage of traffic to another port. They may have a feature to copy 100% of traffic matching a certain filter. Managed switch ASICs have this feature even though it's usually not exposed in the CLI.
Vacuous, like calling a V8 a “next piston firing predictor” because engines are designed so that one piston sets up the next in the firing order and technically there’s some nonzero probability any piston can (mis)fire next. It’s missing two pieces:
1. Useful work that has been done (the previously generated token sequence :: the mechanical work already accomplished)
2. The role of structure in relation to the application (post-training :: other components like crankshaft etc)
A V8 does not "predict" the firing of the next piston, it triggers the firing of the next piston at a precisely controlled time with a spark plug (or a fuel injection nozzle in the case of a diesel engine).
The output of the LLM is literally a probability distribution of what the most likely next token is.
> A V8 does not "predict" the firing of the next piston
It kind of does, though. In a gasoline engine you need to spark the combustion in advance of the piston reaching top dead-center to ignite the fuel early enough that it is able to provide downward pressure on the piston as it rolls over top dead-center. The amount of advance required changes with RPM, fuel octane, etc.
Start of delivery timing in a diesel is similar. You have to do it sufficiently far in advance to account for compressibility of the injection lines, fuel burn rate, etc as a function of RPM. A mechanical governor on an injection pump has a timing advance device built in. Electronically governed injection pumps, or modern common rail systems, do that in software.
So mechanically, engines kind of "predict" the next combustion event. Even moreso when you consider a modern ECU, which may be working at nanosecond resolution to time multiple injection events per cycle. To do this at such a resolution it will have to send signals to components based on a predictive model derived from "past" sensor data. E.g. it needs to act ahead of time to account for electrical and mechanical delays in the system.
Sounds like they blew it. What a shame.
reply