Yeah, this doesn’t appear nice.
Over the weekend, X and xAI proprietor Elon Musk flagged a coming change to his Grok chatbot, by which the Grok improvement crew will take away “politically incorrect, however nonetheless factually true” data from its knowledge banks, as a way to keep away from the app primarily offering solutions that Musk himself doesn’t agree with.
The change has been coming for a while, due X’s Grok chatbot repeatedly offering solutions that counter Musk’s personal opinions on sure subjects. For instance, Grok has advised customers that kids needs to be allowed entry to gender-affirming care, one thing Musk has been a vocal opponent of, whereas it’s additionally countered claims of political violence perpetrated by left-wing supporters (noting that there’s extra proof of right-wing assaults).
Grok has additionally named Musk himself as “the most important spreader of misinformation on X,” amongst its numerous claims which have rankled its creator.
And over the weekend, Musk indicated that he’d had sufficient. After Grok referenced knowledge from Media Issues and Rolling Stone, Musk responded by saying that Grok’s “sourcing is horrible,” and that “solely a really dumb AI would consider MM and RS.”
Musk then adopted that up with a put up calling for X customers to supply examples of “divisive information for Grok coaching,” which has seen X’s viewers present over 100k responses, which they’re hoping can be weeded out of Grok’s data base.
Which is able to make Grok extra aligned with right-wing speaking factors, and extra unaware of factual reporting and proof. In different phrases, it’ll develop into an echo-chamber AI, and with an increasing number of individuals counting on AI for solutions to all types of questions, that looks as if a big concern for broader AI improvement.
Although on steadiness, Grok’s utilization is pretty restricted. ChatGPT reportedly has round 800 million lively customers, whereas Meta not too long ago claimed that Meta AI is probably the most used AI chatbot on the earth, with a billion month-to-month customers.
Grok, by comparability, is barely utilized by a portion of X’s 600 million month-to-month actives, and may solely be totally accessed by paying customers. So it’s not on the identical stage of affect as these different AI apps, besides, the truth that Musk is brazenly stating that he’s enhancing its sources to raised align along with his personal ideology continues to be a priority, notably given current points with Grok’s solutions.
Final month, Grok was discovered to be offering inaccurate solutions concerning the dying toll from the Holocaust, whereas additionally pushing random responses that included references to “white genocide” in South Africa, each of that are primarily based on debunked conspiracy theories.
X claims that each errors had been because of an unauthorized change to Grok’s code by a rogue worker, and that the method has now been up to date to make sure extra checks and balances are in place. However whatever the cause, the incident underlined simply how a lot sway xAI’s programmers can have over the chatbot’s responses, in the event that they so select, and with Elon himself very eager to discredit any sources that don’t agree along with his opinions, that looks as if a harmful combine.
Certainly, the complete xAI challenge was based primarily based on Elon’s personal opposition to different AI fashions, which he believes are being skilled to be “woke.”
As reported by The Washington Publish:
“In an April 2023 interview with Fox Information, Musk stated OpenAI had been ‘coaching the AI to lie’ by incorporating human suggestions that directed the chatbot ‘to not say what the info truly calls for that it say.’ He referred to his new challenge as ‘TruthGPT.’”
So, all alongside, the xAI challenge has been as a lot about political narratives as technological evolution, with Musk seeking to angle his personal AI tasks extra in direction of his personal ideology, reasonably than goal reality, primarily based on web-based sources.
And lots of of his supporters will agree with him. The COVID pandemic has sparked a complete new anti-mainstream media motion, and the truth that the AI instruments are being skilled on what’s thought-about to be mainstream sources is an affront to this push.
As such, Musk’s anti-factual push can be seen positively by many. Even whether it is flawed, even when it results in the expanded unfold of misinformation.
As a result of reality, it appears, is what you make it.
It’s one other instance of the potential negatives of tech advances, which regularly get neglected amid the broader hype.
The identical may be stated of each vital innovation, that whereas there are positives and advantages to be gleaned, we additionally usually overrate the advantages that making such know-how extensively accessible will carry.
The web, for instance, is a revolution in studying, which provides billions of individuals entry to nearly all the data on the earth, which ought to in flip make humanity extra educated, extra knowledgeable, and improve the bottom intelligence stage of intelligence.
But, within the yr 2025, debate about vaccine efficacy, local weather change, and even the very form of the Earth as a planet, are extra vigorous than ever.
Social media was supposed to attach the world, by enabling us to talk with anybody, anyplace, facilitating extra togetherness, empathy and understanding. But it’s arguably performed the other, by offering a way for illiberal, divisive and hateful teams to coalesce, connecting the worst parts of the world. The very algorithms that gas social media engagement incentivize this, and it’s onerous to argue that social media, as an idea, has been a web constructive.
And now we’ve got AI, the subsequent nice hope, which can democratize creativity, and facilitate new ranges of human productiveness, by offering machine-based help to enhance our on a regular basis course of.
How do you assume that’s going to work out in actuality?
Positive, there’ll clearly be advantages, as there was with these different advances, notably from a enterprise perspective. However the Utopian concept that AI goes to usher in a brand new golden age of inventive, clever alternative is actually solely the factor of company pitch decks and boardroom discussions.
Go have a look at the AI content material being shared on-line proper now. Movies with racist undertones that you just couldn’t create with human actors. AI nudes within the likeness of people that haven’t given their permission for such. Folks passing off AI-generated work as their very own, dishonest their manner into unearned alternatives.
This isn’t good, these aren’t good issues which might be being facilitated by this new know-how, and what historical past exhibits us is that the worst parts of society profit simply as considerably, if no more so, from these advances.
Which brings us again to Elon, and his choice to primarily edit historical past to “enhance” his chatbot. That’s a really unhealthy precedent, and a really regarding shift, particularly as an increasing number of individuals depend on these AI instruments for solutions.
Do we actually assume that this can enhance society, or make us smarter as a species?
And if the reply, on both entrance, isn’t any, then why are we pushing AI into each single aspect of each app?
