Why AI may not be as big of a productivity boost as assumed

From The Honest broker : AI Is Defeated in Hollywood—But What About Music? Mr. Gioia writes the same ‘AI vs. the entertainment industry’ article every 2 weeks or so.

I am more of an AI agnostic than either an ‘AI utopian’ or an ‘AI doomer’. I don’t think AI will prove to be nearly as disruptive as pundits predict, whether it’s the entertainment industry or labor. Instead of AI making jobs obsolete, AI will make certain tasks redundant, and will improve efficiency in some respects. Lawyers and doctors are not going anywhere.

The track record of technology actually making a job obsolete or disappear is surprisingly poor. Either the technology complements the job, or the old job finds a niche anyway and is able to coexist with new jobs. For example, it took decades for typewriters to be replaced by personal computers, after the introduction of the first silicon transistor in 1954. Floppy disks still exist and are used. Same for record players. Butter churns, fountain pens, artisan furniture, and horse and buggies still exist too.

The entertainment industry will adapt…like it has always done in the past when confronted by new technology, whether it’s TVs (which competed with radio), audio (which made silent movies obsolete), color (which made black-and-white film obsolete), CDs (which replaced records and cassettes), file sharing (which made CDs obsolete), etc. AI-generated content is no substitute for branding, fandom, and marketing. Top pop artists are still commanding record-high revenue such as touring and merchandise. Geriatric acts are able to extend their careers indefinitely through nostalgia and touring.

Consider Fivrr.com, a popular and publicly traded freelancing site. Revenue has not been hurt by AI offerings such as ChatGPT or Dall-e-2. One would assume that freelancing companies, which tend to specialize in simple, menial online tasks, would be the most negatively affected by AI, but this not been the case. Earnings for Fivrr.com has held steady since 2021 despite GPT, Dall-e, and others:

My explanation is that AI is not as big of a productivity boost as commonly assumed. AI seems to be best at things in which there is no use or need for it. It’s amazing that a computer can play chess better than humans, but there is no use for that. (Even for competitive chess, using computers is considered cheating.) AI can do a lot of things, but productive work is less about being the fastest at domain-specific tasks, but more about merging multiple things to create something new in some new or unexpected way, or manipulating data/content in a certain way that is outside of the scope of the AI software even it it’s otherwise a conceptually simple task.

Let’s assume you need graphics created for a business. Unless your logo or brochure happens to be something like “a basketball player shooting a basket with the Milky Way in the background,” AI is not always faster or easier. Especially when merging graphics with words in such as way that the text is legible and not crowded out by the graphics. If given a choice between spending hours fiddling with Dall-e to get it right, or hiring a freelancer to do it which takes five minutes and $5 instead of hours, the latter sounds like a better value proposition.