Is AI willing to team up with us?
Have you ever seen the 1976 movie “Logan’s run”? In the year 2274, people live a hedonistic lifestyle beneath geodesic domes as part of a utopian society. A supercomputer has the overall control of life and death; everybody’s life is terminated at the age of 30. That seems to be an overly scary outlook on artificial intelligence, doesn't it?
Why worry about that future vision now? Starting as an academic discipline in 1956 and having gone through multiple cycles of optimism ever since, AI has attained significant technical advances, achieving a broad societal presence. This rebirth seems to be a blessing as well as a curse.
On the upside, AI is viewed as a means of making our industries, transportation, communication, safety, security, health and medicine operate more effectively and efficiently. Deep learning can save us a tremendous amount of time, it seems. On the downside, AI substantially adds to the ever-increasing energy footprint of information and communication technologies.
Today, OpenAI’s ChatGPT requires some 3,500 Nvidia HGX A100 servers, consuming an estimated 564 MWh per day, according to research by Forbes. By 2027, this could ramp up to nearly 100 TWh per year, states an article in Joules. This is comparable to the total annual energy consumption of the Netherlands, says Alex de Vries from the Vrije Universiteit Amsterdam. Back in 2015, prior to the current AI boost, a research paper in the journal Challenges reported that ICT could use as much as 51 percent of global electricity and contribute up to 23 percent to the globally released greenhouse gas emissions by 2030 – a scary outlook in times of great concerns about climate change.
By comparison, the human brain only consumes 20 percent of the total body’s energy, according to Simon Laughlin of Cambridge University. And only 10 percent of the brain is in action while the rest remains in stand-by mode. With human power amounting to about 80 watts, this translates to 1 PWh of brain power per year if every human being on Earth is considered. So, the total brain power on Earth is about what we’ll soon invest in AI.
Of course, a well-trained human brain can do much more than AI. Garry Kasparov was a match for IBM’s Deep Blue in 1997 while investing 500-fold less power in chess matches against the supercomputer.
Though highly impressive, that’s not where the human brain works best. Michael Bennet from Northeastern University in the US noted that humans are far better in tasks that require empathy, intuition, imagination and social communication skills, building on 300,000 years of evolution. We often can learn from only a few training samples, whereas AI needs millions. Humans can rely on their gut feeling, while computers are strong in reciting information. We can quickly integrate information from all senses and use integrated perception to make decisions. ChatGPT, so far, can only be approached by text messages. However, multiple AI systems in autonomously driving cars show that AI is catching up fast in that respect.
These characteristics suggest that we shouldn’t be afraid of AI but take advantage and team up. At the same time, we need to preserve and expand on what we’re good at. Education based on extreme strength, discipline and maximum accumulation of knowledge, as promoted by the Chinese-American Amy Chua, the most well-known “tiger mom,” isn’t the way to go. Continuous learning is a constraint, but real success in personal and professional life seems to be nourished by play, music, sports and social interaction, which unleash the skills I mentioned earlier, particularly in childhood.
Unfortunately, room for that is more and more taken over by smartphone apps and social media, which seem to squeeze into any free minute of people’s lives. When riding the streetcar in my town, I see a majority of passengers with headphones, staring at the screens of their mobiles. Empathy, intuition, imagination and social communication skills seem to be delivered by that electronic device rather than being evoked at the human end. That makes me worry that AI isn’t willing to team up with us but rather wants to combat our true strengths. I’m starting to get a little unnerved by the scenario of “Logan’s run” after all.
