Who Does AI Help?

Without glasses, I would be useless. My eyesight is atrocious—absent my thick lenses, the world appears a blurry, Impressionist fog. I sometimes imagine what role I would play in ancient human societies. I’m skinny, clumsy, and have poor coordination. Clearly, I’m not built for labor. The human spirit is indomitable, and I’m sure I’d find ways to muddle through. But unlike my vision, one thing is clear: I would not be where I am today if not for glasses.

Success is contextual and ecological. No trait is intrinsically valuable. Even what we now call mental disorders likely played important evolutionary roles. Many individuals with Autism Spectrum Disorder have above-average memory, perception, and attention, and autistic people may have invented technologies like “oil lamps, multi-component tools, star maps, grinding stones and fire hearths.” ADHD, by causing “impulsiveness, hyperactivity and transient concentration,” might have helped hunter-gatherers survive their constantly shifting, hostile environments. This is the basis for the neurodiversity movement: what society considers deficits are often just differences, and in certain environments, are strengths.

“Every technology has winners and losers,” writes Neil Postman in Technopoly. The printing press helped Martin Luther spread the religion that now bears his name, but devalued the centralized authority of the Catholic Church. Tiny semiconductor chips enabled STEM-obsessed nerds to become billionaires, but hurt jocks: while aggression and physical strength helped our early ancestors hunt and fight, these traits are less useful for writing backend code.

More than any other technology, artificial intelligence (AI) will shape today. As such, it is worth asking: what kind of person will succeed because of AI?

I’ll focus here on the impacts of generative AI, like GPT and DALL-E. I’ll also focus on comparative advantage, relative to the status quo. For instance, while both people with coding skills and without are better off because of GPT making coding easier, I think people without coding skills have a higher comparative benefit from GPT by gaining an entire new skill without having spent years acquiring that skill.

Generalists
I believe generalists gain a comparative edge from AI.

To specialize in something is a bet. The upside potential is the benefit you gain by having a well-developed and rare skill. However, there is also a downside risk if that skill you devoted yourself to becomes less valuable. Because of this, generalists are more automation-proof. Predicting AI progress is hard—you could spend years specializing in, say, predicting protein structures, until one day, a computer does the same thing with near-perfect accuracy. Few predicted how good GPT-4 would be at doing highly skilled, white-collar jobs. Being a generalist is investing in a diversified skill portfolio, hedging against automation.

A second reason generalists are better off is because I think we’re headed towards what I’ll call an “Editor Economy.” In my experience, for tasks beyond a low degree of complexity, GPT usually gets you most of the way there, but rarely all the way there. Reliability will improve as models improve—GPT-4 is miles better than GPT-3.5, for instance—but I doubt that AI will achieve perfection, most of the time. This means, at least in the near future, human oversight of AI output will remain necessary to tie up loose ends, if nothing else. Everyone will be an editor of sorts.

However, to edit requires at least some basic knowledge of the task you are editing. I think this means a “jack of all trades” could benefit more from AI tools compared to “spiky” experts. At my university, you can major in Computer Science, or you can take a mix of computer science and philosophy classes and major in Cognitive Science. I know many CompSci majors who are excellent at coding, but terrible at writing. I also know CogSci majors who are good writers and decent, but not excellent, coders.

If GPT makes a coding error, both majors might find the troubleshooting process equally challenging, given that the majority of the code will likely be fine. However, to edit a GPT-produced text, the CompSci major might struggle compared to the CogSci major. In an Editor Economy, a generalist is better off, more able to refine diverse AI outputs for that last mile.

People With Good Taste
DALL-E and Midjourney have reduced the cost of creating beautiful art to near zero. Despite this, in my opinion, most of the AI-generated art I see online is still… mid. I don’t think this is a limitation of the technology. Rather, it signals a lack of taste.

Visakan Veerasamy has an excellent essay about taste, calling it the “beating heart of all creative value.” He quotes talented people, from Quentin Tarantino, to Paul Graham, to Dorothy Parker, who identify good taste as necessary to excellence in their respective field. But taste is not enough—it usually “has to be coupled with ability of some kind.” As Ira Glass discusses, creative frustration occurs when there is a gap between one’s ability and one’s taste.

Technology narrows the gap. Before the camera, producing a realistic capture of a moment required expert painting ability. After, just a shutter click. Images proliferated after photography, but even now, incredible images remain rare. This is because photography unavoidably involves taste. As Susan Sontag writes, “Even when photographers are most concerned with mirroring reality, they are still haunted by tacit imperatives of taste and conscience.” And good taste remains scarce.

Music production also illustrates this well. Because of software like Ableton or Garageband, anyone can create music, without even knowing what a chord is. Hip-hop producers regularly compose beautiful, virtuosic musical collages, purely by sampling, mixing, and chopping others’ sounds.
AI enables anyone to become an artist, writer, web designer, storybook illustrator, and more. As the cost of production decreases, and as the amount of content converges to infinity, curation and taste become ever more valuable.

Some Personality Traits
Psychologists usually measure personality in terms of the Big 5 Factors:

Again, none of these traits are inherently good or bad—but still, in general, certain traits are more conducive towards success in the modern economy compared to others. A 2021 meta-analysis suggests that Openness, Conscientiousness, and Extraversion are positively associated with higher earnings, while Agreeableness and Neuroticism are negatively associated. This research is correlational and not causal, and the relationships are nuanced. Still, the bigger picture makes intuitive sense. In highly-paid sectors where creativity is prized, like tech, openness is a valuable trait. Conscientiousness, meanwhile, enables individuals to obtain good grades and do diligent work. Neuroticism, linked to anxiety and low self-confidence, is less helpful.

I predict that trait Openness will become even more of a useful personality trait for the AI age. First, technological progress is moving fast, and keeping up with change is easier if you are more interested in new ideas. Second, trait Openness is linked to creativity and intellectual breadth, which is even more useful because of generative AI, for reasons already discussed.

I think conscientiousness, meanwhile, is comparatively less useful now. Discipline is still important for accomplishing long-term goals, of course, but a lazy person can accomplish far more now. Executing projects is easier and quicker, as is learning about new topics. I am not a terribly conscientious person, but AI has made me far more productive.

I am not sure how the other personality traits will interact with AI, if at all. Maybe having AI write your emails in a friendly tone can compensate for low agreeableness? Maybe the relative ease of starting a solo business today will allow more introverted people to succeed? Maybe AI therapy works well, and neurotic people thrive better? I really don’t know.

People With ADHD
I think people with ADHD possess a unique combination of traits that synergize unusually well with AI tools.

ADHD has many downsides. It causes distractibility, lack of focus, difficulties with organization, and more. But AI tools can compensate for many of these downsides. I regularly use GPT to create schedules and to-do lists, combat the blank page problem and “ADHD paralysis,” and automate tedious tasks. I’m not alone: the internet abounds with examples of people using GPT to manage ADHD.

AI also harnesses the positive traits of ADHD. Some studies indicate ADHD is linked to divergent thinking and creative ability. People with ADHD often find themselves overflowing with ideas, but lacking the focus or time to complete all of them.

The bottleneck used to be execution. Ideas are plenty, but implementing those ideas is the difficult part. AI has made execution easier. In an experiment, Ethan Mollick tried to see how far he could go with a business idea in 30 minutes using AI. In the end, it “did market research, created a positioning document, wrote an email campaign, created a website, created a logo and ‘hero shot’ graphic, made a social media campaign for multiple platforms, and scripted and created a video.” This usually would have usually taken hours, if not days, requiring multiple skills and forms of expertise. Implementation is still hard, and Mollick would have to edit those imperfect outputs for any real world use. But with AI, the bottleneck is shifting from implementation to ideas.

This helps people with ADHD, who can turn more ideas into reality and engage different skills in short time periods. I don’t mean to paint ADHD with a broad brush here, or deny the negative, sometimes debilitating, effects it can have on people’s lives. But I do think it could empower many with the condition to live more fulfilling lives.

Conclusion
Predicting technological progress is hard, and predicting downstream effects is even harder. This list is obviously not meant to be exhaustive, and just represents my initial thoughts on who might succeed because of AI. However, I do believe it is likely that an AI-driven world will value creativity, breadth, and individuality even more. (Unless, of course, it causes extinction).

But regardless of who AI ends up benefiting, my broader point remains true. There are no dispositions or skills that are inherently “better” than others; there are no humans who are inherently “better” than others. I hope AI expands the boundaries of our world, enabling more paths for more kinds of people to succeed and create beautiful, interesting, and human things.