HN user
I am convinced that the mass layoffs are occurring because forward-thinking executives can see that AI agents are coming to replace employees faster than most people expect. By doing layoffs early they can cut staff before public opinion shifts to questioning AI as the motive, as opposed to "the economy".
These companies know that AI agents are coming at them hard and fast. I think this and other layoffs are happening now so they can cut down the numbers before the question becomes "Are you replacing them with AI?".
Which tracks with it not impacting "hourly frontline workers employed at the parks and resorts".
"Suing could force the government to start the repayment machinery again, which might not be a terrible thing. Given the low unemployment rate and the existence of income-driven repayment plans for people who are struggling, few people would be ruined by restoring the February 2020 status quo."
This particularly perturbed me with the implication that "ruining" a few people is fine as long as a few other people are made wealthier.
Greatest country etc.
LLMs already have problems with fact vs fiction. I don't see how Reddit of all places has "valuable data" in that regard.
This is great. I'd love this with the same style radio chatter, but from a bot that watches my news sources, tweets, emails, and radios itself to discuss interesting things happening.
"An email just came in at 0743 AM from Steve.. over" ... "Yeah that looks like an interesting followup to the one from John last week. Someone should reply to this one. Over."
I wasn't involved in the legal side and am not a lawyer, but in my limited understanding of US contract law (where this is exclusively used) it is quite simple to create a binding contract that will stand up in court. Building it, if anything, showed how simple the process can actually be.
Funny, we've just replaced DocuSign with a custom-built solution that will achieve full ROI in less than one year, thanks to recent price hikes on the contract.
Great. More e-waste as I toss it in the trash.
Sounds like a fancy way of saying vendor lock-in
The US is doing the same exact thing but it's called 'alignment' and instead of 'censorship' we call it 'safety'.
To stop China we must become China
Or it ends up being both, because Microsoft.
Hope the leaker used good opsec and that this goes nowhere.
Of course it does.
I felt the same way when I really wanted a super soaker 2000 and then the neighbor kid got one before me. "MOOOOM!!!"
Windows 11 is a thinly-veiled state-sponsored malware masquerading as a productivity tool.
Trying to stop the spread of LLaMA with DMCAs is an interesting technique. Like the internet version of trying to put out a grease fire with water.
Since the issue looks to be the consent of the footage being released, would it be the same situation if the footage came from their body cams requested by FOIA? Provided all of their cams didn't simultaneously and mysteriously malfunction of course. Either way I hope things go Afroman's way on this. The raid itself was complete clown shoes.
In practice, the larger the organization the less likely the potential legal sanctions are to dissuade them. My observation has been that once an organization (in the US anyway) grows large enough it is in a special protected status where no real penalties can come to it and there is certainly no risk of exposure to criminal charges for the decision makers.
Source: front page here every single day.
Hopefully there will be a plug-in-your-own-API-key open source thing then. Even better.
I've been following tech long enough to know that as soon as the computer can figure out which button to press it's only going to click on ads, I guarantee it.
I thought to make something similar but what scared me away was the idea that some educational institution could use it and decide a student was being dishonest due to a false positive. I wouldn't want to bear responsibility for something like that.
And secondly, this is a never-ending race. Even if it were to be able to detect ChatGPT content with 100% accuracy today, it would just be used to assist in training another model to defeat it.
I have some Azure services that are not able to consistently make outbound HTTP requests to my heartbeat monitoring service so I'm getting alert after alert this morning. This is just the nudge I needed, and I'll be moving the whole thing to Linode later this afternoon.
The examples provided are not that convincing. Especially the two headlines. Yes they are similar because there are a limited number of ways to ask "Can You Buy a Gift Card With a Credit Card?". Should the topic be off limit now that it's been written about? Take the code equivalent of asking 100 people to write a function in secret that takes 2 ints and returns a sum. I suspect a number of those code blocks will be "plagiarized" from each other too.
Even if they win they lose, because nothing will change. Who is really going to stop using Stable Diffusion at this point?
Like most things Microsoft gets into they're racing straight to third place from the sizable lead OpenAI has established. ChatGPT took the world by storm just 6 weeks ago, being the first large-scale experiment of RLHF, but competitors are biting at their heels already.
From what's being said on social media, the most common complaints are that each update seemingly trades off capability for more safety. A sound strategy if their goal is to eventually integrate it into Teams as Clippy 2.0. Enterprises would be unhappy about a chatbot that actually answered every question honestly. But that strategy will not win them the consumer market.
Large language models at their core are the application of techniques, data, and training time meaning there is virtually no barrier to entry. The techniques are widely known, the data is widely accessible, and a couple million dollars is not unreachable even for individuals. Just like DALL-E is all but forgotten in the footsteps of Stable Diffusion and Midjourney, ChatGPT will soon be eclipsed by free and open models as well.
Consequences. Adams County Sheriff Kimmy Rogers is looking for the word 'consequences'.
Thank you. I have not run into this library before and it looks fantastic.
They may be referring to JA3 [0] which can be used to fingerprint the connection by examining the order of the ciphers sent by client to server during the TLS handshake. It isn't useful to detect that the user is browsing on an iphone, but rather that the useragent says the user is using an iphone but the JA3 gives it away as something else.
It can be defeated by the client sending a specific list of ciphers in a specific order during the TLS handshake but in practice this can be difficult to do as it's typically implemented at a low level. Alternatively, creating a browser extension can also be used to defeat it as the request is running through the browser's code in that case.
Ethics aside, By the time this case is resolved it won't likely matter anymore. I see models getting more granular and specific (and thus smaller) while typical personal computer resources are getting larger. At some point you'll just download the largest 'javascript inference model' that will run on your hardware and be done with it.
There are a huge number of motivated developers that want this to exist, the techniques themselves are not novel, and the code itself cannot be kept from whomever wants to train a model. It's a lost cause.