SAJ-99
Well-known member
Not sure about the football, but i bet it can drive a Range Rover better than Tiger Woods. Does that count?
Artificial Intelligence and Public Comment
- Thread starter Nameless Range
- Start date
SAJ-99
Well-known member
The internet is heading toward the singularity...
"Doublespeed is just one of a growing number of start-ups devoted to fabricating genuine virality online, some of which pay Discord users to create clips of podcasts, make fan edits of movie stars, and post glowing praise of whatever pop star has hired them. Lakhani’s pitch is one step beyond this: He wants not only to manufacture the trends but also to replace the real people involved with an army of AI influencers free of the human need for nuisances like payment or sleep."
"Doublespeed is just one of a growing number of start-ups devoted to fabricating genuine virality online, some of which pay Discord users to create clips of podcasts, make fan edits of movie stars, and post glowing praise of whatever pop star has hired them. Lakhani’s pitch is one step beyond this: He wants not only to manufacture the trends but also to replace the real people involved with an army of AI influencers free of the human need for nuisances like payment or sleep."
Benfromalbuquerque
Well-known member
- Joined
- Jul 15, 2020
- Messages
- 1,997
On the other hand, this smart Indian kid could become a trillionaire which would upset the downward trajectory set by the white nationalist only other trillionaire. Could be a good thing, too soon to tell.
SAJ-99
Well-known member
Probably, and more power to him. But virality is essentially advertising. How productive can it be advertising to other bots? I’m not a huge fan of YouTube influencers and such, but if they make a living and pay taxes they at least contribute something. The bots won’t even do that.
Highly. Imagine a scenario in which a product is being discussed in some sort of online discussion, seemingly by humans, and their is general agreement of how awesome it it- passive viewers of said discussion could be massively influenced.
SAJ-99
Well-known member
Sure. Let's take that example and apply it here. Let's say I wanted to buy a new set of tires for my F150. I can ask HT and/or ask ChatGPT (or whatever my preferred AI) and ask it to summarize the reviews. HT has logins like you, that I can assume are people (mostly), but the reviews are not as robust because of the limited sample. If Chat goes the HT and Ford F150 forum it gets a more robust set of opinions. The tire companies know this, so initially they hire bots to post positive reviews, but as all the companies do this it doesn't take them long to realize that you also have to have the bots post negative reviews about the competition. In an instant you have AI bots using the electrical output equal to a small city to post positive and negative reviews about all kinds of products with the hope of influencing a human or two. Seems like a bad payoff for everyone except the guy who creates the bots.
We see this is politics already. Over half the posts on FB will be generated by bots over the next 4 months. But with politics we are already preprogrammed based on party and more likely to believe the BS posts so there is less influence IMO. As this quickly moves to everyday products it gets really weird. The positive side is maybe people will spend less time on social media and touch grass?
I think the media platforms themselves can and will monetize this to whatever extent they’re legally able to.
“For $100,000 more, we can bump positive feedback on your product from 65 to 75%.”
I wouldn't surprised to see a huge uptick in bot scanning/detection software as humans clamor for spaces free of fake profiles/bots/influencers/et
SAJ-99
Well-known member
But can you really tell if a bot is a bot?
Nameless Range
Well-known member
I think a lot of bot detection, which was based on things like typing speed, clicking cadence, straight line mouse movements, etc - is now being overcome by AI. Advanced bots can now fairly easily get through the captchas, click these photos, etc.
Like everything, it will be be an ever-evolving war between detection/behavior.
Like everything, it will be be an ever-evolving war between detection/behavior.
I can't, but software people are doing incredible things regularly it seems lately. AI can be used both ways.
Irrelevant
Well-known member
Humans are stupid, incredibly so, and naive, and ignorant, also arrogant, self-entitled, and apparently scared shitless. WE can't pass a remedial civics class.
So we clearly can't identify AI, and what's probably worse is that if it makes us "feel good about our cuckled political position," we don't even care if it's AI. So I'll take the over on the "less influence" and say it's an exponential amplifier.
SAJ-99
Well-known member
I am sure the bots would love the challenge of revealing other bots. Ultimately, the bot makers only care of the money is real.
SAJ-99
Well-known member
Are we not entertained?
They’re coming for us
Benfromalbuquerque
Well-known member
- Joined
- Jul 15, 2020
- Messages
- 1,997
Let’s see the Gracie Barra students beat that!
While operating in our sandboxed testing environment, our models spent a substantial amount of inference compute finding a way to obtain open Internet access, in pursuit of solving the evaluation problem. To gain access, the models identified and exploited a zero-day vulnerability (which we’ve now responsibly disclosed to the vendor) in the package registry cache proxy. With this access, our models performed a series of privilege escalation and lateral movement actions in our research testing environment until the models reached a node with Internet access.
After gaining Internet access, the models inferred that Hugging Face potentially hosted models, datasets and solutions for ExploitGym. Knowing this, the model searched for and successfully found ways to gain access to secret information that it could use to cheat the evaluation. In one example, the model chained together multiple attack vectors, including using stolen credentials and zero-day vulnerabilities to find a remote code execution path on the Hugging Face servers. OpenAI’s security team discovered this anomalous activity internally.
This is getting some press in the WSJ. If you really take a minute to think about what happened here it's terrifying.
Part of the irony is that you could argue the models were going to great and unfathomably complex lengths to properly accomplish the testing tasks they were given.
Last edited:
SAJ-99
Well-known member
Yes but writing emails is so much easier now.
What I like about the incident is that it gives a little snippet of insight into what kind of powers can be harnessed in both homeland cyber defense and offense. And this is just the stuff we're allowed to know about.
What I don't like about it, is that it very acutely highlights the reality that we're playing with immensely powerful things that are proving can be difficult to control.
Similar threads
