OpenAI Suspends Developer Behind Dean Phillips Bot 36
theodp writes: OpenAI has banned the developer of a bot that mimicked Democratic White House hopeful Rep. Dean Phillips, the first known instance where the maker of ChatGPT has restricted the use of AI in political campaigns. OpenAI suspended the account of the start-up Delphi, which had been contracted to build Dean.Bot, which could talk to voters in real-time via a website.
"Anyone who builds with our tools must follow our usage policies," a spokesperson for OpenAI said in a statement shared with Axios on Sunday. "We recently removed a developer account that was knowingly violating our API usage policies which disallow political campaigning, or impersonating an individual without consent." OpenAI apparently is not a fan of Richard Stallman's 'freedom 0' tenet, which argues software users should have the freedom to run programs as they wish, in order to do what they wish (Stallman is careful to note this freedom doesn't make one exempt from laws).
The suspension and subsequent bot removal occurred ahead of Tuesday's New Hampshire primary, where Phillips continues his long-shot presidential bid against President Biden.
"Anyone who builds with our tools must follow our usage policies," a spokesperson for OpenAI said in a statement shared with Axios on Sunday. "We recently removed a developer account that was knowingly violating our API usage policies which disallow political campaigning, or impersonating an individual without consent." OpenAI apparently is not a fan of Richard Stallman's 'freedom 0' tenet, which argues software users should have the freedom to run programs as they wish, in order to do what they wish (Stallman is careful to note this freedom doesn't make one exempt from laws).
The suspension and subsequent bot removal occurred ahead of Tuesday's New Hampshire primary, where Phillips continues his long-shot presidential bid against President Biden.
Pssst... (Score:3, Informative)
... don't tell these people about all the open models (including quite capable ones) that anyone can run on their own computers! ;)
Re: (Score:3)
That doesn't really have anything to do with it, does it? They have a policy, it was published and available. The developer violated it, so they terminated the relationship. Everything seems in order.
Re: (Score:2, Insightful)
Re: (Score:3)
Interesting take. How do you feel about "Truth Social"?
Re: (Score:1)
That's different. As president the Oompa Loompa was banned by the MSN from getting his message out. There was nowhere else for him to turn to. If the corrupt deep state run out of the basement of that pizza joint in D.C. and funded by Soros hadn't barred him speaking, this wouldn't be needed.
Re: (Score:3)
I understand why it's there. I'm drawing the parallel to the use of the word "truth" here. Truth Social is a place you retreat to in order to specifically not be called out for being untruthful.
Re: (Score:2)
The point is that if you run your own models, you don't have to care about someone else's policies.
Beyond the fact that it's also a lot cheaper, too, and you have infinite customization possibilities.
Re: (Score:2)
That might've even been an intentional move to generate more news about Dean Phillips' candidacy, first with the bot announcement, then with the ban.
Re: (Score:2)
But why? (Score:2)
I'm sure the AI version was in no way inferior but heaps cheape...
Oh. Right. I forgot that it cannot mimic the most important feature of a politician: How do you bribe an AI?
Re: (Score:2)
I don't see the issue with opening up a Patreon, Gofundme or other similar account in the name of an AI :-)
Re: (Score:2)
How do you bribe an AI?
I thought big tech were the briber and not bribee.
Re: (Score:2)
They've come full circle now, instead of buying politicians, they just roll their own. Cut out the middle man and all that.
Ranked Choice Voting (Score:5, Insightful)
Ranked Choice Voting now: You vote for who you actually like. If your 1st choice does not make it your vote is not 'wasted', your vote transfers to your 2nd choice.
Re: (Score:2)
Most likely this time you're voting for the vice-president or the handler.
Re: (Score:3)
Single transferable vote is awful. It tends to elect the most extreme people that have high support among people in the political extremes rather than people with broad bases of moderate support. If just over 1/3rd of the electorate are Nazis, just over 1/3rd are Maoists and just under 1/3rd normal people who just want a normal life without "revolution" or "cleansing", and everyone is happy with the moderate candidate, but the Nazis would still prefer a Nazi (even though 2/3rds of the electorate would hate
Re: (Score:2)
Sounds like a nice, clean, defensible strategy. And there is no way it would not break down into CHAOS when the last vote is cast. Certainly not in the U.S.. Completely impossible, save perhaps for small, inconsequential races. You can't even get people to align on declaring the majority vote the winner - what makes you think this non-intuitive method would result in people nodding their heads and agreeing, "That's fair."?
Re: (Score:2)
What on Earth are you talking about? Vote order is irrelevant. Nothing happens until polls close.
"In the election between Andy Politician, Betty Frontrunner, and Cindy Candidate, Betty was ranked higher by voters than both Andy and Cindy - therefore, she was elected!" Man, so confus
Re: (Score:2)
You missed my point entirely. You will not be able to "sell" it to the people. What you think is patently obvious will be rejected outright the second their preferred candidate loses because of the nuance you relish.
Your leap to objecting to "the last vote is cast" is a little annoying. You fixated on the literal meaning instead of the intent. Technically you're right. Conversationally, you're out of bounds.
Re: (Score:2)
I'll repeat:
"In the election between Andy Politician, Betty Frontrunner, and Cindy Candidate, Betty was ranked higher by voters than both Andy and Cindy - therefore, she was elected!" Man, so confusing, how could people ever understand that...
Re: Ranked Choice Voting (Score:3)
Everything you said about STV is semi-random nonsense. Are you sure you weren't taking about plurality?
Secondly, STV is for multi-seat elections. Condorcet is for single seat. It's like comparing a screwdriver with a spanner
That's centrist nonsense resulting in far-right... (Score:2)
A great example of uselessness of game theory in real life situations though.
For one, it equates the choices - long-term and short-term, conservative and progressive, moral and ethical, economic and social...
E.g. There wasn't a WW3 with China OR Soviet Union spurred on by a myth of genetic superiority and bloodlines built-into the leftist ideology. Right-wing ideology will always boil down to such a myth.
For two, it assumes a whole bunch of other things, not just the choices being essentially identical.
From
Re: (Score:2)
And here I am without mod points!
Fortunately, this hasn't yet been down-modded to "Troll."
non-impersonator (Score:2)
Where's my choice of unabashedly AI candidate ? not an imposter but rather a machine generated candidate that is marketed as a machine and is algorithmically bound to live up to its pre-election promises when in power ?
Give me 50 to choose from and let them all make different sets of promises
maybe limit it to just one or two representatives initially (eg t represent DC for the first time) and see how it goes
Re: (Score:2)
Can't get on ballots. I remember many years ago when someone tried to run "Ficus For President". Failed because they couldn't prove it was a natural born citizen of the US, of the required age.
You could however get a person to pledge to abide by all the decisions made by an AI.
everything (Score:2)
Everything not compulsory is prohibited. Welcome to the new free state.
Someone at OpenAI supports Phillips... (Score:2)
I'm sure they were probably just protecting their terms of use, but it's hard to think that someone isn't intentionally sparking a Streisand Effect here. "Kicking" the Phillips "campaign" off your platform and releasing it to the news just draws attention to Dean Phillips while giving a counter-example to "big tech is in the DNC's pocket!!1!1!1!!!eleventyone!!1!"
Nobody cared that a dead-end long-shot campaign that nobody outside of the political news obsessed even knew existed was using generative AI to an
violation of terms, not of spirit (Score:4, Informative)
The summary confusingly left out an important point: the creation of the bot was contracted by a super PAC that is supporting Dean Phillips' candidacy. They wanted to use the technology to promote their guy.
But OpenAI has policies against use of their platform for political reasons, presumably to avoid deception by malicious actors, which goes against the presumption here.
So the report isn't about someone doing something bad and misleading people with AI, as many of the Slashdot articles seem to be these days, but, rather, about a company enforcing their TOS. Understanding that fact, in turn, kind of makes it a non-story.
Re: (Score:2)
So the report isn't about someone doing something bad and misleading people with AI, as many of the Slashdot articles seem to be these days, but, rather, about a company enforcing their TOS. Understanding that fact, in turn, kind of makes it a non-story.
Wish I had some mod points. Well said. If you are going to use proprietary software, you have to play by their rules. Lots of other LLMs are available (LLaMA and BLOOM come to mind) that have no TOS to get in the way of your use case.
That is unless (Score:2)
wtf (Score:3)
"OpenAI apparently is not a fan of Richard Stallman's 'freedom 0' tenet, which argues software users should have the freedom to run programs as they wish, in order to do what they wish (Stallman is careful to note this freedom doesn't make one exempt from laws)."
That paragraph debunks itself.
That freedom doesn't make one exempt from laws, and impersonating someone is that when others could reasonably take you seriously.
Taking it in (Score:2)
I suspect that will remarkably change when the party being simulated is a different one.
OpenAI provides a service, not software (Score:2)
OpenAI apparently is not a fan of Richard Stallman's 'freedom 0' tenet, which argues software users should have the freedom to run programs as they wish, in order to do what they wish
OpenAI provides a *service*, not *software*, so Stallman's 'freedom 0' tenet has nothing to do with it.