Now this is where I completely agree with you! Sanders is taking a dead-wrong approach on AI: make laws to prevent its use. His concern (even though he might not ever confess to it…) is that we need to do that to protect jobs. You can’t do that.
But by the admission of the AI monopolies, completely unregulated use can/will be dangerous. So did I suggest outlawing it? Hell no! I just said: they need to carry insurance or escrow a bond for mishaps. Once damage has been done to consumers, will they be able to face off against a tech monopoly in civil court to recover for the damages? What happens if the company dips?? (<joke>in the case of OpenAI, that’s now looking more possible!</joke>) The monopolies want a windfall when they sell; they don’t want to know anything when they were party to damages whose paid product helped abet…
My opinion of AI coding agents is not based on what I’ve read or what I seen on youtube. My opinion is based on my own personal experience. I have a claude code account I’ve been using daily since I signed up last February. In that time I’ve build about a dozen projects that I use daily.
One of them is an ios app that I’ve configured to send every photo/video I take to a folder in my home server. I basically completely replaces icloud: GitHub - priestc/media-backup · GitHub
Ironically, I’m using AI to get away from using big tech products, because I’m building my own version. Not everyone will do this. Most people will just use spotify rather than build their own spotify. But now it’s not just big companies with lots of resources who can make something like spotify exist. Any person with an idea can make something a reality. We no longer have to wait for big tech to build stuff for us.
Also, AI is not a monopoly. You can run AI models on your own computer and not give a dime to any tech corporation if you want.
Yes; if you check my posts I mention the self hosting alternatives. And I always concede when AI tech has something to offer (provided it’s not hosted by a model provider that charges up the wazoo!) I do not agree with Sanders’ approach on AI.
But Anthropic, OpenAI, Google, Meta -they ARE monopolies! (Anthropic being the least egregious of the lot…)
Wow, Patrie, you are very, very well educated and informed. It’s almost like Cenk has a “burner” on this discussion platform. Some of this stuff is so exhausting. Does AI need immediate and aggressive oversight? Obviously, the industry needs significant civilian oversight. For me it falls into the same category of issue as does American campaign finance. What is there to discuss? What reasonable, lucid, critical thinking person thinks it’s a good idea for man to create God? By the same token who thinks virtual legalized bribery of the entire body politic is a good move? Maybe Socrates was right about democracy? If there is a better example of why he did not care for democracy, than the rise of Trump, I cannot think of it. Why do we have to sit here and treat him like a serious person just because he tricked enough people (with a lot of help from fucking Hollywood) into thinking he is a “businessman” and will run the country like a business. Well, in fairness, they were right. He is running America like he did his business affairs. Everything is on fire and he is sucking whatever value out of the business that he can. Then leave he leaves the bankrupted carcass for dead. I am not one who thinks this mess just spontaneously appeared, and I know it’s not all his fault, but the point stands. At any rate, I don’t know how Cenk does it. He gets up there night after night and tries to explain to the world what so CLEARLY is happening, yet nothing fucking seems to get through. Israel too falls into this bucket of stupid bullshit we shouldn’t have to “debate” with unserious, corrupted, donor proxies who go by job titles like news anchor, senator, etc.
Well, it seems pretty straightforward to me. By their own admission, they need safeguards. Safeguards to me means liability insurance; and if you can’t get the insurance, you need to be bonded (escrow money as dictated by regulations, to be able to reimburse for future losses…)
The Bro’s are hoping that such regulations will get tied up in Congress. FTS[1]! Do it at the State level!!
Check out their lack of an answer when asked what insurance coverage they have, in the post above (as reported by that communist news outlet FORBES!)
“Self regulate”? Talk is cheap; show us your insurance or bond that will cover future damages, even if you end up in default…
Don’t let AI corporations speak for the entire technology. There is more to AI than just the 3 or 4 corporations that always gets talked about on the news.
I don’t think there needs to be “safeguards”. AI is just software, it already has enough safeguards.
If an AI “does harm” it’s because someone was using insecure software. AI can only exploit a vulnerability, if one exists in the first place.
If you want regulation, then make it illegal to use insecure software. Then if an AI agent hacks the Australian government, then you press charges against the Australian government for being in violation of the new regulation that makes running insecure software illegal.
I don’t want to make it illegal to make any software! Except if it’s software that was designed to operate Walter White’s robot machine gun (if it happens to be used on, for example, school children that is…)…or to access a biohazardous lab for deadly pathogen formulas…etc. Not so much for the agent implementers, unless they are doing willful destruction like that.
There was a great discussion on “Pivot to AI”; they pointed out how a few years ago people who did takedowns like Lulzsec would get extradited and prosecuted for that. But now that monopoly labs did the same thing? “Woopsie!”
No, no: you f’d up; you pay damages for that. You take the bad with the good profits! The insurance/bonding thing just makes sure they will. They can try anything that is not purposely destructive; but if they are reckless and cause damage -they have to make good on their damages. You have a freedom to create, but coupled with that is a responsibility to not be reckless, wiring up models or agents to facilities that can do harm. Pre-agent, making sure software you made was not harmful was on every programmer’s checklist. Now with agents: “hell let’s wire it up and see what happens, without any forethought to what might happen…”
Sanders wants to outlaw AI; I disagree. But hugely valuated companies already demonstrating their recklessness? Get thee to an insurance broker (or banker, to escrow funds for damages…)
States can make that happen. If a State wants to be completely laissez faire, and damages are done? The politicos in that State can answer for that. If other States have a light touch and they recoup for losses? They can be praised. We’re going to learn as we go; but I don’t think we should be waiting for losses to occur before we even set up a safety net for reimbursing for losses. That’s the entire Tech Bro’ accelerationist line of thinking: “Apres moi, le deluge! I get to sit things out in New Zealand or Argentina!”
The idea that an AI can cause “damages” is just absurd.
Its like saying people who draw dangerous looking cartoon characters should also have to have an insurance policy in case one of those cartoon characters “breaks loose” and causes havoc on society. It’s just so patently absurd. A drawing on paper will never ever ever ever come to life and break out of the paper it was drawn on because that’s not how reality works.
An AI “breaking out of it’s sandbox” is equally as absurd, but only to people who know what a sandbox even is. Nearly 100% of the people in media that reported on this issue have no idea what a sandbox is, so when the liars at OpenAI told everyone we should be scared of this happening, everyone went long with it.
The only “damage” an AI can do is access information put on the public internet that the person who put it there did not intend to be accessed. The solution is to not have stuff accessible on the open internet that you don’t want to be accessed.
Its like if I have valuables in a shed behind my house and I don’t put a lock on that shed, AND I place a sign in my front yard that reads “unsecured valuables in unlocked shed around back->”. It’s not anyone else’s fault by my own when those valuables get stolen.
Then why are the largest model producers and Gates saying that??
I myself say it’s absurd. But can you USE AI resources to perpetrate damages? (which is what the reality is…) Sure! Software malpractice is as old a story as the Therac-25.