Artificial intelligence will reach human levels by around 2029. Follow that out further to, say, 2045, we will have multiplied the intelligence, the human biological machine intelligence of our civilization a billion-fold.
What is AI – is a perplexing question that deals with a seemingly enigmatic phenomenon. John Haugeland, in his book Artificial Intelligence The Very Idea notes that research in the field of AI is not merely about mimicking intelligence to produce a clever fake. Instead, AI is about machines with minds. That is a startling description for AI. The concept of a “mind” has tantalised philosophers for centuries; nevertheless, if we are to believe the hype around AI, it would appear artificial minds are indeed being engineered. Is it really?
Haugeland puckishly notes that when it comes to AI, people can be categorized into two camps – “scoffers” and “boosters”. Scoffers find the whole idea of AI propitious and ridiculous. On the other hand, boosters believe that artificial minds are inevitable. Unless we get some of the basics of AI technology right, we could inadvertently find ourselves in the wrong camp.
Types of AI
AI can be divided into two categories: narrow and general. Narrow AI is often used synonymously for weak AI. Weak AI is about testing hypotheses about minds but would not be a mind. Weak AI systems appear, behaviourally, to engage in intelligent human-like thought but in reality, enjoy no form of consciousness. Weak AI systems at best can provide aids to human thought. Raymond Kurzweil, in his book – The Singularity is Near, describes weak AI as “Artificial intelligence that performs a useful and specific function that once required human intelligence to perform”. This definition sits well with examples of Weak AI, such a Roomba vacuum cleaner (they are brilliant by the way), Siri, and self-driving cars etc. An important distinction to grasp, is that every AI system in the world today is almost certainly a Weak AI system. Such systems are designed, programmed and trained to carry out only a specialised task.
General AI, on the other hand, is human-level intelligence and with broader application. A term usually associated with Genera AI is Artificial General Intelligence (AGI) for scenarios when AI is applied to problems in many different domains, similar to how human intelligence can. Another synonymous term with AGI is Strong AI. Kurzweil defines Strong AI as “artificial intelligence that exceeds human intelligence”. It’s a computer system that is capable of becoming a mind – Intelligence as smart as you and me. AGI systems can execute tasks it has never been programmed to do after figuring out by itself how to perform such a task. There are no real-world examples for AGI. There is not even agreement on how to build it – or even if it is possible.
So, there you have it, the hype about AI that you hear in the news is most undoubtedly Weak AI, and it is not going to take over the work. At least not yet!
Machine intelligence is the last invention that humanity will ever need to make.
So, does that leave you in the “scoffers” or “boosters” camp? The likes of Elon Musk, Grady Booch and Stephen Hawking, I think, sit firmly in the “boosters” camp raising voices of concern against AI in crescendo, drawing urgent attention to the potential dangers of AI. In an open letter, they state: “because of the great potential of AI, it is important to research how to reap its benefits while avoiding potential pitfalls”. If you want to read it, it’s here.
I’ll leave you with one last thought. In September 2020 (yes last month) a college kid used OpenAI’s GPT-3 language generator to generate blogs and watched in surprise as several of them went viral. All that the kid had to do was provide the AI system with a headline and introduction for the post. GPT-3 returned a full article. AI is now writing blog posts and we humans are unable to tell the difference. The Massachusetts Institute of Technology’s Technology Review magazine reported (see) that one of these posts also reached the number-one spot on the popular Silicon Valley website Hacker News! Surely, that says a lot on how much AI technology has advanced in recent years.
