Sam Altman is seeking $7 trillion (!) to develop a new computer chip to help speed up "advanced AI" development. Is it worth it? Who will pay for it?
Just less than a year ago, prominent scientists wrote a letter to call for a global halt on AI development that has since been signed by over 33K people around the world. Guess what? Not much changed, except the major AI hype wore off. The "intelligent" chatbot that spurred industrial strikes and awe is now mostly a brainstorming partner and ghost writer, while new AI models continue to be developed. For factual and specialist knowledge, we are still better off asking human experts.
What the letter really symbolises is a soft democratic protest against a monopoly of tech institutes. Radical change demands radical measures, as Extinction Rebellionists would surely agree with. A complete switch off was an unrealistic plea, though. There simply is too much skin in this AI game - albeit not so artificial anymore. In fact, the blurry boundaries of a "metaverse" are slowly dawning upon us.
My doctoral research touched on the impact of generative AI on our sense of what is real and true. Multiple studies (by others) find that people cannot distinguish AI-generated from authentic human-generated content anymore. This was shown for text and images. Realistic AI-generated footage of celebrities and politicians on social media and porn sites can attest to this. However, in spite of the gloomy prospect of unfettered use of generative AI (as many such AI models are freely available and/or open-sourced), I caught myself in the brief naivety of thinking these are new problems.
Whatever form of AI is not fundamentally going to change what we, humans, already do - fakery not exempt. Deceit is as old as human civilisation. "The Art of War" is a military strategy book written by Chinese general Sun Tzu more than 2000 ago. Modern-day deceit exists in many walks of life: from romance scams (think the "Tinder swindler") and business fraudsters (think Elisabeth Holmes), to marketing and entertainment. Great actors are great liars. Products are embellished in advertisements. Pop singers get autotuned. Princes have innate privileges, for they are "god given". Of course humans will use AI to invigorate their fantasies, because we love imagining a reality that transcends the human condition - and we all buy into it, form artificial girlfriends to action movies.
On that somewhat perverted note, I recalled an audience question during the Alan Turing Institute's cybersecurity conference last year which inevitably discussed AI in the domain of cyber defense:
"If we ever achieve artificial general intelligence (AGI; AI that matches human intelligence), should we treat it with the same law?"
I agree with the principle that any intelligible being should abide by the laws that were intended to be applied to such entities. If AGI, a step further from AI controlled by humans, would do anything against the law and we hold it accountable for it has human-level intelligence, why should we not judge its behaviour by the same standards we apply to ourselves? What repercussions would make effective deterrents to such AGI, however, leaves reason to invest in further research...
Although I may seem late to weigh in on the real issues around AI (was not invited by Rishi Sunak, weep weep - but also see my previous posts on AI), this may be just the right time to ask deeper questions about those deeply intelligent models that mister Altman so eagerly wants dibs on. How is "advanced AI" going to be more than a supernerd fantasy? By what standards are military applications thereof justified? What is the tech bro community's actual objective function?
It might help to regard A(G)I as a catalyst and not some divine end in itself, contrary to what some tech hot shots may believe. In this vein, I found Demis Hassabis' UCL Prize Lecture last year greatly inspiring for its reasonability. He talked about plenty examples to show how AI is already functioning as a catalyst for scientific breakthroughs. Better healthcare through personalised medicine and reducing climate change are only a few pressing real-world challenges that may not be solved without AI.
Disclaimer: my writings are not enhanced, nor consulted for with AI.
Comments