When the Godfather of AI is Afraid of his baby.
As the debate surrounding the threats of artificial intelligence (AI) has escalated to the global agenda with cases of abuses skyrocketing following the release of ChatGPT in November, the warning against the technology and resignation of AI Godfather Geoffrey Hinton from Google in May further added to questions whether the world is going in the right direction.
Who is Geoffrey Hinton? And why should we heed his warning? Apart from the comment that it is a more urgent threat than climate change, the fact that a part of him regrets his life’s work should suffice.
Hinton is among almost 28,000 tech leaders, scientists, and AI developers who signed an open letter calling for a “pause” of giant AI experiments beyond the systems more powerful than GPT-4, which is the latest version of AI and much smarter than ChatGPT. Other signatories are Tesla founder Elon Musk and Apple co-founder Steve Wozniak, for example.
Known as the godfather of AI, the 75-year-old BritishCanadian cognitive scientist and computer scientist is the brain behind the advent of the near-human-like intelligence, building his career on development of neural networks decades before the technology went mainstream. The networks consist of complex computer models whose structures mimic the human brain.
While age is part of the reason, freedom to discuss the dangers of AI without considering how this impacts Google is the main reason, following cases of abusive uses of AI which sparked debate on how it should evolve as the machine has become increasingly sophisticated.
“I wouldn’t like to devalue climate change. I wouldn’t like to say, ‘You shouldn’t worry about climate change.’ That’s a huge risk too. But I think this might end up being more urgent,” Hinton said, noting that while it’s clear what to do to control carbon emissions, that isn’t the case for the ultrasmart thinking creation.
Hinton’s passion for intelligent machine started in his childhood. He is the great-great-grandson of the mathematician and educator Mary Everest Boole and her husband, the logician George Boole, whose work became one of the foundations of modern computer science. He graduated with a bachelor of arts in experimental psychology from King’s College in Cambridge, and did PhD research on artificial intelligence at the University of Edinburgh in 1978.
After decades working at universities and institutes, he joined Google in March 2013 when his company, DNNResearch Inc. was acquired, resulting in his two jobs as a university professor and corporate researcher.
At that time, Hinton’s dreams to create ‘a ghost in the shell’ was seen by the majority in scientific community as a joke. For him, there was a lot more to ‘deep learning’ by the machine than processing words and numbers.
In his interview with MIT Technology Review, his unique views stemmed from having a biologist father. “So I was thinking in biological terms. And symbolic reasoning is clearly not at the core of biological intelligence.” What he believed was true: the neural network can be rewired by changing the sets of coding which are programmed to have learning ability.
“Crows can solve puzzles, and they don’t have language. They’re not doing it by storing strings of symbols and manipulating them. They’re doing it by changing the strengths of connections between neurons in their brain,” explained the man who revolutionized the AI world twice with his work on neural network and deep learning. The breakthrough earned him the Turing Award in 2018 (also seen as the Nobel Prize of computing).
Hinton believes AI intelligence will surpass that of humans in every field from art to science although human supervision is still a must at some points such as the input and finalization processes.
Compared to tech giants like Microsoft and Google, he can be quite pessimistic. But if you look at it with biological lens, it’s not unreasonable. Artificial General Intelligence (AGI), he says, will be the next frontier of AI technology, which is exactly what PaLM 2 is.
Days after his departure from Google, the company launched PaLM 2, a “next-generation language model” which it says outperforms other AI systems in many tasks. It will be infused with 25 new Google products and features to enhance user experiences from search results and photo taking/editing to writing computer code or a job application letter, thanks to its ability to search, collect, analyze all the data to give users the best answer and solution.
Although the working fundamentals is the same, the cutting-edge AI version is so advanced that it requires very little human involvement. By simply typing a phrase or keyword, PaLM 2 has proven it can deliver a near-perfect answer to us in less than a minute.
What a stellar performance…But everything comes with a price.
When GPS and Google Map were launched years ago, it was touted as one of the most celebrated digital tools that would save our time, gasoline money, and ultimately our brain energy for more important tasks. That problem is, not only does this kind of technology simplify our lives, but also our brain’s structure. Translation: less brain exercise leading to loss of brain neurons and intelligence.
Studies have shown dependence on GPS and Google Maps have a detrimental impact on the hippocampus, an area of the brain which plays a major role in learning and memory. The hippocampus is also the first that gets hit by Alzheimer’s disease.
Neuroscientists have been studied how AI influences human brain and mind. They have found that regular use of Siri or Google weakens the brain’s ability to retrieve information, resulting in impaired thinking process as we forget information or where we can get it. In 2011, a group of researchers from Harvard University had named it ‘The Google Effect’.
Whether Hinton’s fear of AI overthrowing human race in the way depicted in sci-fi fiction will come true has remained unknown. But if it ever happened, the first step would certainly be robbing human intelligence.
Then again, his statement on climate change might become real before that. With current frequency and severity of climate change phenomenon, humanity might perish long before AI dominates the planet, particularly when the need for massive electricity is added to the equation.
Since the booming of AI, power consumption has been exploding. At present, it is unclear how much exactly a powerful AI system like ChatGPT demands at peak performance.
According to a research paper published in 2021, training GPT-3 took 1.287 gigawatt hours or about as much electricity as 120 US homes consume in a year. This equates 502 tons of carbon emissions. Another way to measure it is to look at Google’s electricity bill, 10-15% of its total electricity consumption comes from AI. Annually, the company’s AI burns around 2.3 terawatt hours, about as much electricity as all the homes in a city the size of Atlanta, the research paper found. In general, the thinking machine gobbles up more energy than other forms of computing. A single training of AI consumes electricity used by more than 100 US homes in one year.
Hinton’s resignation is a huge loss to Google, but a boon to the society. While artificial intelligence is the reflection of our own intelligence, it also poses an intellectual and mental challenge…how to enrich our life with technology while maintaining physical and mental health.