I’m a lawyer and listening to this call was absolutely painful. I like to think I’m a pretty decent negotiator, and agree that there was likely a few million here were this handled competently.
Don’t be afraid to bring people in when something is outside your area of expertise.
Man, I have a totally opposite view about LLMs expressing creator’s values. Not only do they express them, they don’t STOP expressing them to the point of utter annoyance. Any remotely PG topic ends with a safety caveat, e.g., “however, it’s important to consider . . . .”
Those statements don't come from the base model, they come from the steering methods, basically a form of hand tuning after the model is mostly trained, which the paper says are relatively crude and imperfect. It is the fact that these models are so unpredictable that led to these attempts at steering.
Reminds me of the early Bing where you could talk to it for much longer periods of time. You could get it in a state where it would give some terrible reply and then an upper layer would delete that message and be like "oops, didn't mean to say that"
Observational hypothesis based on 5 minutes with the app (and asking it to repeat the prompt it was given):
This does not use any OpenAI service to analyze the image. Rather, it’s using some other image, recognition model to generate a description of the image, feeding that description into chatGPT, and having chatGPT hallucinate other likely details based on that description.
Aside from the question of whether this plan was legal, DoNotPay seems like a terrible product. The results it generates seem laughably bad, and it’s questionable whether “AI” is actually involved when it takes them literal days to generate a document for certain types of requests.
https://www.techdirt.com/2023/01/24/the-worlds-first-robot-l...
Indeed - and this is not new. Many years ago, I took a look to see what all the fuss was all about.
From start to end, he/his product seemed amateurish. From giving out a herokuapp.com subdomain in early press releases (which were republished on major sites), that was then no longer in use (allowing it to be taken over), through to the actual generated output.
When I looked at a letter it generated, it was laughable. The "chat bot" was simply asking questions, and capturing the (verbatim) response, and throwing it straight into the template. No sanity checking, no consistency, etc. There was absolutely no conversational ability in the "chat bot" - it was the equivalent of a "my first program" Hello World app, asking you your name, then greeting you by name.
It wasn't capable of chat, conversation, or comprehension. Anything you entered was blindly copied out into the resulting letter. Seems nothing has changed.
Most of these things seem to be hybrids, humans overseeing automation, with varying degrees of human involvement. Guessing they at least have a review queue for non-boilerplate docs about to go out.
Note that the Samsung Internet privacy policy is awful. They collect:
“The information we obtain in this manner may include, identifiers associated with your devices, types of devices, web browser characteristics, device and operating system type and characteristics, language preferences, clickstream data, your interactions with Samsung Internet (such as the web pages you visit, links you click and features you use), dates and times of your use of Samsung Internet, and other information about your use of Samsung Internet.”
Does the house have coax wiring for cable tv? If so, get a pair of MoCA Ethernet adapters. I have 2.5Gbps ones running across a large house and they are rocksolid reliable.
She’s representing a client and putting up what she thinks is the most effective way to argue their case. Amazon will put up whatever they think their best defense is. A court will decide who is correct.
I personally found the repetitive "no household uses" to be a weak and sensational argument. There are any number of uses for something like this for someone who might have a hobby involving these chemicals.
Plus, it's not like there aren't many things _with_ household uses that can be used for suicide too.
I love Rock Auto. The way they make it easy to see which warehouse your items are shipping from and choose other items from the same warehouse is super cool too.
Just ordered something from them earlier this week. They got my repeat business because the last time I ordered something from them it was effortless (other than their kinda poor UI lol) and the item arrived super fast. I guess the shipping was a bit pricy, but that's fine. They had the part and got it to me promptly, and charged a fair price for it. Works for me!
https://www.reuters.com/legal/us-supreme-court-lets-metas-wh...