the_mar

the open weight llms didn’t just happen to be as an act of god. someone created them and released them. i don’t see why they should not be liable.

that depends how you define hacking. i am not a lawyer, but i think there’s an element of some sort of gain (data, money), otherwise if a tree falls in the forest…

i think you misunderstand what i said. as much as i dislike ai ceos, i doubt anyone is setting out to build ai with explicit intent to do crime. they are just not doing a great job preventing it. just because they know it is in the training data, does not prove intent

why do people need to be 10x more productive?

personally, i want a 2 hour 3 martini business lunch, not to ship more slop at all hours of the day

idk abt you but i just replaced a faucet by watching a youtube video of a middle aged guy. ya’ll are using opus 5.5 for that?

i have non-tech friends who actually do personal software. personally i don’t have any desire to. i don’t think grandma would justify the valuation tho

what you mean why? most companies with spare cash payed 10-50x what they actually need in tokens, just to see how it goes. it’s useful, it’s not that useful, so while the price per token can increase, everyone is getting smarter about spend. and even price has a limit, you can’t make ai more expensive than a human, because then everyone would just hire humans.

idk what is this you people do, but most of my day is not spent on writing code. most of my day is spent in meetings abt budgets, timelines, evaluations, compliance and marketing. i don’t think ai is going to automate corporate circus anytime soon

you know what i miss? when slack didn’t exist. when it was perfectly acceptable to wait 2 business days to respond to someones fuckass email

i think when people say today’s prices they do not mean subscription tier, they mean the enterprise paying 10M+

as a person, i would absolutely not pay more than $1K for ai, because at that point i can just get myself a bartening job

but the argument here is that the ai company does not explicitly train agents to hack, it’s that it doesn’t prevent them from doing so sufficiently. so it’s negligence and not malicious intent

you generally need a license to operate a vehicle, and a governing body ensures you know applicable rules and laws and are capable of driving. so you can’t say “i was unaware that hitting a person with a car was going to kill them” or “i didn’t know i could not drive on the sidewalk” or “i didn’t know where the brake pedal was”.

you can say that asking an agent to “get you a spot in a class”, does not classify as deliberately committing a crime. a user cannot know that the agent will resort to hacking as opposed to refreshing the reservation page at some rate in case spots open up

im not sure that’s true, if a human agent does something illegal, it’s not always corporate liability. an individual can be liable too, e.g. if a pharmacist deliberately gives you cyanide instead of aspirin, that person has committed a crime. if a pharmacy doesn’t have proper protocols for medication storage and identification, and it happened by mistake then a pharmacy is liable.

i think a person can be liable for improper use of agents if they explicitly and deliberately used it for illegal purposes (aka intent). i don’t think it matters if it was openclaw or muse. i think a company can be also liable for giving access to unsafe product to consumers, but it does not absolve the human from responsibility

but it doesn’t auto-translate into “because your dog did it, you are not responsible”. not that agents are similar to pets. your dog is trained by you presumably, but an agent a regular person has much less understanding of than their dog.

i think ultimately the company providing said agents should be responsible for not educating their customers and not providing sufficient guardrails.

afaik hacking is illegal, we don’t have specific legal framework for agents per se, but if we say agent is a tool human instructs and operates, then a human is responsible. and if the tool operated outside of it’s stated capabilities or is otherwise faulty, then a manufacturer is responsible, no?

what do you mean “say what you will about climate change”

giving out unrestricted control that is. in the example of Waymo, a human can control some aspects of the car manually, but it can never override low level obstacle detection or say open the trunk/door when the car is moving.

because we all can agree child porn is bad and just because a company donates to a charity doesn't mean they get to dictate their policies

they banned energy shot drinks?!

there is no cloud, it's always someone else's computer, so technically someone who is physically in google/apple datacenter at the exact moment your data is being operated on, hypothetically would have a chance to steal your data.

saying that because google owns the hardware, they would quietly not delete or copy your data is quite a wild accusation.n

pardon my ignorance, but why does compute hardware pose any security concerns?

From the information presented, the privacy case is not that your data is only accessible to you (which arguably can have a backdoor) but that the data is NOT stored at all, so it's not possible to build a backdoor. I know there are probably other ways around it, but it's my understanding is that no data is kept on any server when the response is sent back to your device

well, name an example of a thing that can never change then.

"research alternatives" meaning what exactly? You think open source is somehow not susceptible to the same issue, plus all of the malicious updates?

I for the life of me could not solve the <18 example from wikipedia. but the number/color one is super easy

by your definition, what is marketing then?

it included flights, hotels, food and travel expenses for 9000 for multiple days, as well as the "party". US-based travel for 1 person for 5 days is easily 4K, on top of that some people were probably international so it would be higher, and on top of that there are the "party" expenses like venue and catering which probably wasn't that significant.

it's a very simplistic take. the issue with ChatGPT is that it speaks with authority, vs webMD and such just provide information. to say that how the information is presented is irrelevant to the outcomes is reductionist at best

a friend of mine was a creative director and a big tech co until recently, she was replaced by AI

do you genuinely think that numerous meetups isn't a marketing push?

I mean, do you think no one at OAI and every other lab has vibe-coded some agentic demo? The problem (?) is that when you work at a corporate job you have to think about security.

It was killer compared to alternatives. All other "homepages" of the internet were the cluttered mess of ads.

I feel like we are arguing semantics though. But IMO any UI that does the job that consumers want well is good UI. Just because it was simple doesn't mean it wasn't good