The first artificial intelligence (AI) system was a robotic mouse that could find its way out of a labyrinth, built by Claude Shannon in 1950.
The technology has come a long way since, explains an expert, as he retraces the brief history of computers and AI to see what we can expect for the future.
We now have AI systems like DALL-E and PaLM with abilities to produce photorealistic images, and interpret and generate language
But, by 2040, there will likely be a transformative AI system that can match the capabilities of the human brain, he predicts.
Despite their brief history, computers and AI have fundamentally changed what we see, what we know, and what we do. Little is as important for the future of the world, and our own lives, as how this history continues.
To see what the future might look like it is often helpful to study our history. This is what I will do in this article. I retrace the brief history of computers and artificial intelligence to see what we can expect for the future.
Since the early days of this history, some computer scientists have strived to make machines as intelligent as humans. The next timeline shows some of the notable artificial intelligence (AI) systems and describes what they were capable of.
The first system I mention is the Theseus. It was built by Claude Shannon in 1950 and was a remote-controlled mouse that was able to find its way out of a labyrinth and could remember its course. In seven decades the abilities of artificial intelligence have come a long way.
Capabilities of AI systems are now comparable
The chart presented displays the progress of artificial intelligence (AI) over the last two decades. The data shown is based on tests conducted in five different areas, including handwriting recognition and language understanding, where both human and AI performance were evaluated.
To make the data comparable, within each of the five domains, AI’s initial performance was set at -100, and human performance was used as the baseline, set at zero. This means that when the AI’s performance crosses the zero line, it indicates that the AI system scored more points in the relevant test than humans who did the same test.
Until a decade ago, AI systems were incapable of providing language or image recognition at a human level. However, as the chart shows, AI systems have become increasingly capable over the years and are now outperforming humans in tests across all five domains.
However, it is important to note that the performance of these AI systems in real-world scenarios is mixed. In some cases, these systems are still lagging behind humans. In contrast, some AI implementations have become so cheap that they are available on the phone in your pocket, for example, image recognition that categorizes your photos and speech recognition that transcribes your dictation.
AI is here Now
The passage explains that rapid advances in AI technology have made it possible for machines to be used in various domains. Some examples of these applications include:
- AI systems being used to determine flight prices, monitor passenger behavior at airports, and assist pilots in flying planes.
- AI systems being used to determine loan eligibility, welfare, and employment decisions, and even who gets released from jail.
- Governments purchasing autonomous weapons systems for warfare and using AI for surveillance and oppression.
- AI systems programming software and translating texts, with virtual assistants now commonly found in households.
- AI systems making progress on difficult scientific problems.
- Large AI systems called recommender systems determining what we see on social media, what products are shown to us in online shops, and what gets recommended to us on YouTube.
The author notes that AI technology is no longer just a futuristic concept but is already impacting all of us in various ways. The broad range of applications means that AI can be used for both positive and negative purposes. Therefore, it is important for everyone to develop an understanding of AI technology and how it should be used in the future.
Finally, the author speculates on what AI technology might be capable of in the future, given the rapid progress made in just the last two decades.
What is next?
The text explains the history and evolution of AI systems over the past eight decades. The chart provided in the text plots the relationship between the amount of computation used for training AI systems and the year in which the AI system was built. The chart shows that as training computation has increased over time, AI systems have become more powerful.
Training computation is measured in FLOP, which is equivalent to one addition, subtraction, multiplication, or division of two decimal numbers. This measure of computation is essential for machine learning-based AI systems as they require training to improve their capabilities continually.
The chart is plotted on a logarithmic scale, where each grid-line represents a 100-fold increase in training computation. This long-run perspective shows a continuous increase in training computation, which has been in line with Moore’s Law for the first six decades, doubling roughly every 20 months. Since 2010, the exponential growth in training computation has accelerated, doubling every six months.
The text highlights that AI systems such as DALL-E and PaLM, which can produce photorealistic images and interpret and generate language, respectively, are among the most powerful AI systems built to date, requiring the largest amount of training computation.
The text concludes by posing a question about what we can learn from the historical development of AI for the future of AI. The exponential growth in training computation suggests that AI systems will continue to become more powerful, and it is essential to consider the ethical implications of such technology as it is increasingly integrated into our lives.
