Winters and summers of AI and is this time different?

The algorithms and models on which today's AI developments are based have a long history.

In fact, since the beginning of modern computing, in the late 1940s, scientists have looked at computers as machines to become ?intelligent”. The definition of intelligence however has not been created.

Described in a 1951 article, the famous 'Turing test' assumed that an 'intelligent' program was one that gave a human the impression that it was dealing with another human being.?

However, the Turing article did not specify, for example, for how long such an impression should last. Would today’s ChatGPT pass it? At first- of course. But as we progress to try it make mistake??

The coining of the name of artificial intelligence is credited to John McCarthy in 1956 during seminar of computer scientists in Dartmouth College. The name was not very much liked, but McCarthy wanted to distinguish ?intelligent computers” from the general field of ?cybernetics”. ?I had to call it something, so I called it artificial intelligence”.

The Dartmouth seminar is a special moment in the development of computer science and can perhaps be seen as the beginning of the development of the discipline so named by McCarthy.

When it comes to the history of AI, it's important to recognize that the field has experienced both summers and winters. Summers are periods of rapid progress and breakthroughs, while winters are periods of slow progress and stagnation.

In the 1960s Herbert Simon, a future Nobel laureate, predicted that ?in 20 years machines will be capable to do any job a human can do”. Where do we know it from?

In the same period, Marvin Minsky, founder of the MIT AI Lab predicted that ?within generation problems of creating artificial intelligence will be substantially solved”.

Currently, we are undoubtedly in a summer for AI. In recent years, AI has made significant strides forward, with breakthroughs in machine learning and deep learning leading to new applications and use cases. However, it's worth noting that this isn't the first summer that AI has experienced.

Back in the early 1970s and later in the 1980s, AI experienced a summer of its own, with rapid progress and excitement about the possibilities of the technology. However, that summer eventually gave way to a winter, as progress slowed and funding dried up.

In that time Science Research Council in the UK and the Department of Defense in the US issued very negative reports on the progress of AI. With this week’s letter of Elon Musk &Co- does it ring a bell?

So, the question is, is this time different??

For several reasons, it seems so.

Firstly, there is the post-2010 gigantic increase in computing capacity as a result of the cloud revolution. The computing power now available is astronomical compared to what the Dartmouth Seminary used to have.

Secondly, the explosion in the 2000s of digital photography, the social media revolution and smartphones has meant that humanity has produced more data in twenty years than in the previous few thousand years. Commentators have dubbed this phenomenon 'big data’. And big data is being used for training alghoritms.

Thirdly and finally, after almost 60 years of various attempts to approach the AI problem (symbolic AI, subsymbolic AI, expert systems), the key concept of AI has become machine learning and deep learning based on it. Large neural networks are capable of creating large generative models. ChatGPT has probably been played with by all of us.

The current summer of AI has prompted much discussion and speculation about the emergence of 'general AI'-an artificial intelligence that will be equal to humans. Ray Kurzweil or Mark Zuckerberg have spoken out in favour of the imminent achievement of AGI. But again- there is the question of the definition of intelligence.

The position of Stephen Wolfram, distinguished professor of computer science and AI researcher, seems very interesting here. A computer program is intelligent if it is able to compute on its own.?

ChatGPT cannot do this, so according to Wolfram, intelligent is not.

Prominent cognitive scientist Steven Pinker has stated that AGI is always 25 years ahead of us, as it always has been.

And will the current summer of AI last forever? Today we would say yes. Yet, 30 years ago computer scientists who worked previously on AI projects tried to conceal it in their CVs.

However, as John McCarthy-"AI was harder than we thought" stated 50 years after the Dartmouth conference.

Video recommendations this week:

I recommend book that helped me to write this article: