Can machines think? What is consciousness? These questions are too important to leave to AI researchers. Cognitive scientists, psychologists, philosophers, and neuroscientists have investigated these questions for ages and came up with frameworks to think about such questions. Here we focus on two things, the philosophical critique of AI (both symbolic and connectionist) and various approaches to consciousness (or intelligence, or minds).

Recently, a Google engineer came to the conclusion that LaMDA (a language model) is sentient. Another Googler published an essay in The Economist in which he is contemplating the future of AI and when large language models become sentient. The first guy has been sent on temporary leave by Google, the second one is a vice-president and a leading figure at the company. Both guys suggest that LaMDA is sentient, conscious, or at least intelligent. As Emily Bender writes, human-like programs abuse our empathy – even Google engineers aren’t immune. These technologies are far from being intelligent, we still don’t fully understand their effect on us, yet big tech introduces them to the market. Think of The Social Dilemma, and you don’t want that bright AI-driven future right now.

The philosophical critique of AI

AI has often been thought of as a kind of science that tries to reproduce or even outperformed human problem-solving capabilities or as a kind of engineering discipline that doesn’t pay attention to biological and/or psychological theories when dealing with automating tasks that are handled by humans. AI researchers always think there will be a huge breakthrough within a decade or so. E.g. these days Geoffry Hinton thinks that deep learning will be able to do everything. History teaches us to take such claims with a grain of salt. Almost 60 years old ELIZA seems to be as intelligent as today’s big language model based chatbots. Although it can imitate the behavior of a Rogerian psychologist, no one would call it intelligent. Just have a look at some of the rules it employs below, and you’ll see why.

ELIZA and the symbolic AI tradition were on the wrong track as Dreyfus’ seminal book brilliantly showed us. Almost all of Dreyfus’ critiques were realized as true ones by AI researchers. After publishing What Computers Can’t Do, the philosopher became a pariah in the AI community because he questioned all the underlying assumptions of symbolic AI. Later the very same man was seen as a forerunner of connectionism, postphenomenology, and embodied cognition.

Gary Marcus plays a similar role these days. The cognitive scientist (and author of great pop-sci books, successful startupper, etc.) got a reputation as a criticist of the connectionist movement (which later became the deep learning school). The Algebraic Mind is a superb book on the problems of connectionism and an attempt to integrate it with symbolic approaches. Of course, Marcus has been derided by deep learning researchers, but he has got a special sense of humor and he loves playing the protagonist of a different but not exactly defined approach. Although almost everyone says that this time we are about to make a breakthrough in AI, we can hear similar claims since the inception of the field.

We propose that a 2-month, 10-man study of artificial intelligence be carried out during the summer of 1956 at Dartmouth College in Hanover, New Hampshire. The study is to proceed on the basis of the conjecture that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it. An attempt will be made to find how to make machines use language, form abstractions and concepts, solve kinds of problems now reserved for humans, and improve themselves. We think that a significant advance can be made in one or more of these problems if a carefully selected group of scientists work on it together for a summer.

Wikipedia, Dartmouth Workshop

What is this thing called the conscious mind?

Although his thoughts are controversial, Dennett’s books are great sources if someone is interested in the evolution of minds. From Bacteria to Bach and Back is not just a vivid argumentation for naturalism, it also collects valid questions about minds, consciousness, and evolution. Although Dennett concludes the question of artificial minds with a positive answer he notes that even though they are theoretically possible, this doesn’t mean they can be built in practice. Even if they can be built in practice, we are far from the technological sophistication required to do it so.

There is a tradition in the philosophy of mind and other cognitive sciences, which was very close to connectionism. The theory of the embodied mind holds that we are not just brains in a vat, but living organisms with bodies. We are situated in an environment that stimulates our sensors and our sensing has an effect on our minds. The theory of extended mind goes a step further and states that we use external tools in our environment to augment our cognition (think of places that help us to memorize routes, but books our even our mobile phones extend our mind). Andy Clark’s books are excellent sources to learn more about these things and how this is related to the theory of distributed representation (aka connectionism).

Do we have answers?

Sorry, but no, we don’t have final answers. Having ultimate answers to the big questions of life is the feature of religions and ideologies. Surely, we made lots of progress since the inception of modern science, we just think that science cannot stop at one point having reached a state where every possible question has been asked and answered.