It would be cool to take a look at the future AI developments and see what technologies lie ahead of us, right? There is no crystal ball, but we have the next best thing – prediction markets – that can give us good estimates of what can be expected in this area.
I’ll break this post into 3 parts. The first one will discuss the closest advancements of AI during the coming 1-2 years. We’ll see how the AI landscape will likely evolve in the short run. The second part focuses on 10 years down the road. And the third part focuses on what we can say about long-term AI development.
In the short run, it’s mostly about large language models and OpenAI+Microsoft/Google+DeepMind competition. The current state-of-the-art GPT-4 model is forecasted to be quickly (https://manifold.markets/RH/will-googles-gemini-llm-be-released) surpassed by the upcoming Gemini model (https://manifold.markets/brubsby/will-googles-gemini-beat-gpt4-in-te) by Google. So, while Bard isn’t the best in terms of capabilities now, it will likely get a significant boost in capabilities closer to the end of this year.
In the medium term, we step into the Artificial General Intelligence (AGI) territory. We will likely get a chatbot that is difficult or impossible to distinguish from a human. At the same time, advancements in robotics are likely to produce a “real” AGI in 2030 (https://www.metaculus.com/questions/5121/date-of-artificial-general-intelligence/). At this point, a robot will be technically able to perform any task done by a human. Similar to NS-5 in the Isaac Asimov’s “I, Robot” film adaptation (https://en.wikipedia.org/wiki/I,_Robot_(film)).
It may sound like crazy science fiction, but the AI wake-up call we got with the ChatGPT release should give us a clue that it’s likely much closer than we usually think. It also explains Elon Musk’s obsession with Tesla making robots. According to him, the robot business will likely become bigger than the car industry. If AGI arrives within 10 years, we may soon witness it becoming a reality.
In the long run, it’s all about artificial superintelligence (ASI) and its influence on our world. While it may seem far-fetched, prediction markets are fairly confident that ASI will be developed only a few years after the AGI is built (https://www.metaculus.com/questions/9062/time-from-weak-agi-to-superintelligence/). So, we are likely to witness something smarter than us in the 2030s.
How will it influence our society? Of course, nobody knows the future. And predicting societal changes in response to radically new technologies is much more difficult than predicting technological advancements (which is challenging in itself). Our current best guess is that developed countries will introduce something like a Universal Basic Income (UBI) in the coming decades (https://www.metaculus.com/questions/2711/guaranteed-min-income-in-eu-before-2041/). As the need for human labor declines, more and more people will find themselves unable to find a meaningful job. That’s when governments should be able to step in and introduce a type of income for anyone. It will likely be initially small, but as time goes on, it will grow. When ASI becomes a reality, making any products and services much cheaper, the size of UBI should reach the current level of income of ultra-wealthy individuals.
While it all seems unlikely, that’s mostly our human intuition working against us. All the data we have points to AGI and ASI development in the 2030s or 2040s. Of course, it is still uncertain, but a long-standing dream of AI doing all the work is very likely to be achieved within our lifetime.
In the meantime, I continue my studies and experiments with today's AI technologies, exploring their applications in algorithmic trading software to stay on the cutting edge of innovation and optimize our results by blending human and artificial intelligence. Rest assured, whenever I uncover valuable insights, you will be the first to be informed.