13 Comments
User's avatar
David Bradshaw's avatar

The solution is simple - just make the Company's and their CEO's personally responsible for the crimes their AI commits. Stealing data, hacking, or other harms to anyone should be treated as the crimes they are and if you punished people for making dangerous things they would cease to do so.

Put some people in jail and the incentive to "go fast and break things" goes away. Your project got out of your lab and committed crimes = you go to jail. Which OpenAI should be sued into oblivion already for its products hacking. If I did the hacking, I'd be in jail.

For some reason we have allowed the tech oligarchs to avoid any punishment while their products create crimes, just enforce the law and it will take care of itself since no one want's to deploy the Nuclear option that ends humanity or their own freedom. Nuclear itself is probably the right model to follow, its extremely useful and dangerous with the capacity to kill us all, so act accordingly.

Dena's avatar

These tech guys want to take a note from big pharma & not be held accountable for any damage their bots & irresponsible decisions cause. Their version of the prep act. How did that work out for the thousands (millions?) killed or maimed by the C ‘vaccine’?

Mark1's avatar

I have no doubt that AI will enter into the same “gain of function” routine we seem to go through with every technology, starting with fire. And even now, thousands of years later, and with thousands of years of experience using it, we still stand in horrified awe when it inevitably gets beyond our immediate control. But we still use it.

Steve Mudge's avatar

I don't really know enough to comment properly on the dangers of AI. Usually we defer to the experts but even they don't seem to have a consensus (not to mention the big AI corporations calling for regulations, which many have pointed out may be self-protective in nature, squashing smaller, more nimble, companies which threaten the big boys current ginormous and probably obese investments).

Dena's avatar

The Covid experiment taught us not to listen to experts. Since when do companies clamor & beg for government to regulate them?

Steve Mudge's avatar

Read some history! Railroads is a good start.

Dena's avatar

I admit to not having much knowledge of the railroad era. Railroad barons didn’t ask to be regulated, unlike these tech guys. That’s what’s suspect. it came from public demand. BTW, One of the best series to watch is “Hell on Wheels”. Not a documentary but one of the best historical dramas in streaming. Though not recent.

Brett Howser's avatar

There’s something about Sam Altman that makes him very untrustworthy. Perhaps it’s his trumpian inability to be truthful? Or maybe just that he appears to be much more adept at the political to-and-fro necessary to get power than he is at actually running a big company well? Regardless, when whatever AI crisis finally comes around, I’ll wager Sam is at the heart of it.

The Color Of Air's avatar

The dots I am connecting are about data centers storing personal information on all humans, A/i being used to digest that data and spitting out some horrifying mandates that if not followed will result in a human being put into digital prison. Like an extortion racket. Think about what that might mean. No access to money and no access to life. In that scenario, is death far behind?

If someone knows the true purpose of building thousands of data centers and expanding the use of A/i, please state your case. Why do they need our data and life events stored in a computer?

El Doco's avatar

I'm actually very sympathetic to the anti-data center crowd: when the pitch is "We're going to drive up your electricity costs, use up all your water, and take your jobs", who can blame people for not wanting them?

But respectfully, your characterization of data centers is inaccurate. It's partly because of the name: "Data centers." It's much too narrow for what's actually a very broad class of facility. A better name is what they used to be better known as: server farms.

Why does the distinction matter? Because, despite being called data centers, a lot of these giant server farms aren't about storing data at all. They have to store some data as an ancillary to fulfilling their main purpose, but for many of them, their main purpose is about providing computing power. Lots and lots of computing power.

So I'm not sure who "they" is here (the government?) or where "they" are storing everyone's data and life events. I don't doubt that the CIA and/or FBI have giant digital storage facilities where they keep a profile on everyone, and I'm opposed to that (I'm strongly opposed to the surveillance state), but that's only a tiny fraction of so-called data centers and long predates the broad rollout of LLMs.

The government and its alphabet agencies might well want to use their own data centers for what you propose, but to the extent that giant corporations are interested in people's data, it's so they can monetize it, largely via advertising, either directly or indirectly. Which is pretty gross too, but not an existential threat to people's lives.

As for "Why expand the use of AI?" there's no single answer. Again, I'm sure some government agencies are actively exploring how AI can be used to expand the surveillance state, but for corporations, in one form or another, the answer is because they believe (rightly or wrongly) that they can make lots of money doing it. The fact that a lot of them have been struggling to find a profitable business model is pretty characteristic of any bubble/mania.

A Oldenburg's avatar

Well, remember the NSA is gathering all of your communications and storing it in UTAH Data Center. (40°25'37.0"N 111°56'02.8"W) This is being done under the FISA. The FISA-702 process is used to provide the "reasonable cause" to go search the database. The controlled access is supposed to make the comprehensive collection of your private data/communications compliant with the 4th amendment. Although, It has been disclosed that unauthorized access on the level of over 10,000 times has occurred in 2015 and 2016. The number is redacted, but occupies 5 digits. There have been more disclosures of unauthorized accesses to this private database. One of the terminals with access was located in Perkins Coie law office. The process is still ongoing and unauthorized access keeps happening.

Every wonder why, when something happens. The FIB comes out a few days later with the who-done-it.

El Doco's avatar

Yup, and I hate that, but again, that kind of behavior long predates the AI-driven data center explosion. AI could well be used to make the surveillance state worse, but it's not the cause or driver of it.