The music is too loud. Uri Levine gets up, he doesn’t want to talk here. Just around the corner is a table behind a glass door. There, the well-known entrepreneur can explain in peace what so many start-ups and corporations are doing wrong when they rely on artificial intelligence (AI). AI currently dominates the tech industry. At least that part of the tech industry that meets like every spring these days for the SXSW conference in Austin, Texas. Everyone is talking about the chatbots, in which AI fools users into answering like a human. And it’s much more than hype: AI will change the tech industry like social networks and smartphones.
For example, thanks to chat AI, can any one-person start-up now offer 24/7 customer service because the voice robot can answer questions at any time? Or can only the large tech companies finance the necessary AI computing power and seize even more power on the Internet? These are all questions that are on the minds of the industry – and the people here in Austin – right now.
Levine shows how much artificial intelligence can influence everyday life. If you avoid a traffic jam with Google Maps, you can thank him. He is co-founder of the company Waze, which was bought by Google. Waze has offered AI navigation. To say that exactly, Levine would find nonsense. This technical description is completely irrelevant to the users, he says at the quiet table behind the glass door. What matters to them is the following offer: “I’ll help you avoid traffic jams.” That solves a real problem, people download a new app for it.
Too many companies wouldn’t look at the problems plaguing users. You would fall in love with technical solutions. That’s what’s happening with AI right now. Many founders and financiers see technology as a hammer with which they want to hit all kinds of nails – regardless of whether the nails should be hammered off or not. “Such hype only creates value for the start-up investors, not for the customers.” Levine demands that successful start-ups have to think the other way around when it comes to hammer and nail. You would have to identify a real problem people have and really figure out what bothers people about it and why. Only then is it a question of which technical solution is suitable. Levine has also just published his mantra as a book in English (“Fall in Love with the Problem, Not the Solution”).
The AI race between big and small will be decided in the next two or three years, says influential management consultant and futurist Amy Webb. “It’s going really fast now and we’re not prepared.” She fears that more and more artificial intelligence could trigger a tendency toward monopoly. Few corporations win – and consumers get nothing.
No more people browsing the internet, “the internet is browsing you”
Also, artificial intelligence’s thirst for data is dramatically changing the Internet, says Webb, whose talk was one of the best-attended at the Austin conference. So far, people have gone online when they are looking for something: the way to the café, new sneakers, the news. The rampant AI is now turning that around, Webb said. Because these machines work better the more information they have tapped. Collecting data is becoming more lucrative than ever. So it’s no longer people who search the Internet, says Webb, but the other way around: “The Internet searches you.”
Entrepreneur Levine is pessimistic that established industrial groups will be able to make big leaps with AI. Because in order to develop a completely new business model, the company would have to stand up and say: “What we are doing is wrong.” But corporations are too scared for that, which prevents innovations. If you really want to make a breakthrough, you have to invest seven to 10 years, says Levine. His anti-congestion company Waze was founded in 2007 and bought by Google in 2013 for more than $1 billion. Corporations tend not to be able to hold out for such periods, says Levine. Because in these years the AI project is not yet profitable. And then a bad fiscal year intervened, which is why the board unfortunately, unfortunately, has to bury the AI project, Levine predicts. That’s it for the innovation.
You can have a lot of fun with AI, says the head of the company that invented Chat-GPT
Greg Brockman is one who has used AI to drive innovation – so much so that everyone is talking about it now. But right now he seems tense. He has clenched both hands into fists and placed them on his thighs, which looks a bit cramped. The head of Open AI should explain himself on stage. His company brought Chat-GPT to the world. Hardly any product has conquered the Internet as quickly as this chatbot, which writes love poems on command, fills gaps in the code for computer programs, and delivers marketing ideas as bullet points ready for presentation. Great, say some. Dangerous, say the others. Because not every idea of the AI robot is clever, not every sentence written into the world by Chat-GPT is covered by reality.
Brockman explains how such an AI works. First, she reads an incredible amount of text, all sorts of things: bad texts too. Also wrong texts. “That’s the base model,” says Brockman, and it includes “every prejudice, every ideology” that people have formulated. But then comes the second step: the chatbot is brought into line. Humans would intervene in the AI’s calculations, making the chatbot more likely to spit out something sensible. To show how much manual work this is for the programmers, Brockmann raises both hands while explaining this and pushes an invisible mass forward as if the basic model were a large lump of dough that is then kneaded well until the finished AI is served.
You can have a lot of fun with AI. Brockman cites “Game of Thrones” as an example, the very popular television series with the rather unpopular ending. Anyone who was annoyed by the finale can have a new ending written with AI, he says. Or you can have yourself written into the series as a star, and the AI can then visualize the new episode. Brockman is diplomatic about possible problems with AI: Of course, AI is not perfect. He would also welcome new laws that regulate AI. But the benefits would outweigh it. For example, AI can take over content moderation in social networks better than humans. These moderators decide, for example, whether an uploaded photo may not be shown for legal and ethical reasons. Or whether a comment is just completely outrageous or constitutes hate speech that should be deleted.
People should not overestimate AI
Do such sensitive decisions really belong in the hands of machines? Esther Perel warns against overestimating AI. Perel is a psychotherapist and has a popular podcast about love, sex and partnership.
A major flaw of the AI from their point of view: Despite the large amount of data, artificial intelligence often lacks a concrete context. For example, in therapy sessions like Perel’s: she plays an audio recording on stage with a Ukrainian father. His wife fled the country and misses him. And she misses compliments from him. He can’t understand it: people are dying in Ukraine – and his wife can’t find the words “You look good today”? According to Perel, the underlying problem of this couple’s dispute can only be identified with human empathy and context: the man is suppressing all his feelings because of the cruel war. Unfortunately also for his wife.
People should not overestimate AI, warns Perel. Getting married, moving away from the family, having a child or not, opening the relationship to third parties, having a ventilator switched off – there are no technical solutions for such big life and love decisions.