Видео

Thank You - John is an incredible technologist!

Great observations. Here is some more official details for those that are interested: www.cisa.gov/news-events/news/attack-colonial-pipeline-what-weve-learned-what-weve-done-over-past-two-years www.law.georgetown.edu/environmental-law-review/blog/cybersecurity-policy-responses-to-the-colonial-pipeline-ransomware-attack/

They happen to be legally binding. You can read the case that happened with Air Canada. Search for an article named "What Air Canada Lost In ‘Remarkable’ Lying AI Chatbot Case"

Ironically I was going to mention the same case. As an update; the Judge ruled that the Airline must honor the agreement it made with the man, because they were the ones that "employed" the bot, & therefore it should have known the rules, provisions etc, but just because it didn't they can't say it wasn't our responsibility. He likened it akin to a parent's child- stealing something from a store, & the parent claiming they weren't their parent. These things are not Mutually Exclusive.

A dealer is a dealer is a dealer right...? 😅

Soooo make a hackathon and test your bots, but to be honest every bot is exploitable

Do you use your own LLM or take a current LLM and modify it to your rules?

Depends on the client, but we do both@@BySixa

nah not ours haha@@mm.786

@@BySixa depends on the business needs, we do both

Hire humans to do your customer service. Easy

​@@dalehendricks8201 that was not the questions and not helpful. I seen couple aproches i don't know what is correct aproch but: 1 input sterylization and flaging when Somone tryies to break bot 2 i know that on trening phase you can (try?) to fight with jailbroking by giving more empasis on trainibg data and Lea on current conversation But ultimetly the problem is that chat box just completes sentences (quesses next characters) not talking with you

@@incognitoideveloper9608 although those solutions very well might work, I think a much simpler and more robust approach would be to just update the websites terms of services. Why hire a team of software engineers to fix/ maintain the software when you could just hire a lawyer to revise the TOS to state that the chatbot cannot provide a legally binding contract.

Sentient AI maybe? JK lol but seriously the best solution is going to be found in the training side of things. In some way you have to account for pretty much every possibility. Right now the big race is trying to complete a list of what those negative possibilities are and how to train them out of the system.

My personal thoughts here are, education, knowing when to use these new technologies in your business, and practical regulation. If you are wanting to leverage AI/ML in your business, understand the risks before diving in 💡 It'll be a few years before the "excitement" and "energy" start flattening, and the real value applications, that are secure and can be operationalized, start making it to market.

Likewise. You are in that category too with USNUA 🔥

Approx. 12 degrees on the VRAM.