Is an AI Backlash Imminent?
Corporate leadership loves AI for its cost savings. Consumers, less so.
Part One of Two
Every year or two there is a new buzzword in technology:
“Blockchain”
“Web 3.0”
“Influencer Marketing”
How can you tell these are buzzwords? They become passé as quickly as they become ubiquitous. If you’re stuck with the wrong signifiers in your pitch deck or sales campaign, you can kiss that deal goodbye.
This spring, Google (GOOG) and OpenAI introduced AI assistants for consumers, but it’s not clear that anyone other than college students too lazy to plagiarize their own term papers really wants their help. Despite the fanfare, the real B2B advantage comes from cost savings — in other words, cutting jobs.
Which jobs? Most typically, customer service and content production roles.
Zoom (ZM) and Visa (V) waxed frothy about the potential of AI in a closed-door Goldman Sachs conference this week; meanwhile Nvidia (NVDA) is betting the size of the total AI market opportunity will reach $600 billion.
For perspective, the global online entertainment market (streaming movies, music, gaming, videoconferencing, social media, and reviews and ratings site) was worth $367 billion in 2022. The current market cap of Bitcoin, still the most highly valued cryptocurrency is “only” $500B.
"Down the road, AI will help you schedule meetings," Zoom CEO Eric Yuan remarked, "Essentially, we're looking at collaborative AI."
But every meeting scheduled by an AI means one more human with less to do. In other words, more office managers and receptionists on the unemployment rolls.
“Retraining for the AI economy” is not as glamorous as it sounds; most of the actual work takes place for sweatshop wages in the developing world.
From a macroeconomic standpoint, this is bad news — and we haven’t even talked about harm from AI’s carbon footprint, first exposed by Timnit Gebru, or from bias in big data profiling systems (as detailed in a feature in last month’s Rolling Stone).
The scope of different AI technologies — from predictive to generative to large language models — is so vast that it’s impossible to generalize. It’s impossible to see the downside of work carried out by the late Karen Bakker, who used machine learning to better understand the meaning of animal sounds and pave the way for interspecies communication.
Interactions with LaMDA-like systems can feel magical and wonderful. Computer scientists even have a name for the unexpected behavior of AI models; they call it “emergence.” Much to my surprise, last summer I was able to “teach” a Project December chatbot to play blackjack in about 20 minutes — no coding or training data required. I just held up the cards and told the AI what they were.
From a financial perspective, the blueprint for evaluating AI’s long term relevance as a technology is quite simple and non-mystical. Do AI tools deliver a better result than the equivalent human workflows, or are they just cheaper?
What analysts need are better benchmarks, and more transparency about when and where AI is used. And yes,
“There’s a prompt for that.”
Results in Part Two.