aivarsjurcans.com

Design by Artis Briedis

The ‘wow’ factor we experienced when first seeing and trying the ChatGPT two years ago would be difficult to match. Nevertheless, I  am very curious about what will be the next breakthrough and when we can expect this to happen.  

These are my three takeaways from watching several videos on the direction that future artificial intelligence may take featuring some of the biggest names in the AI space:

Lesson #1: The slower, the wiser

The latest versions of large language models already claim that they are becoming significantly better at reasoning. Instead of giving instant answers, they are taking 30-60 seconds to think before answering which means considerable progress in moving towards System 2 thinking.

(according to Daniel Kahneman, System 1 thinking is fast, automatic and intuitive; it is based on patterns and experiences. The System 2 thinking is slow, deliberate and conscious, and is being used for complex problem-solving and analytical tasks.)

This mode of thinking is expected to enable models to form a hypothesis, use the knowledge it has accumulated to refute or affirm it, and continue reasoning from there.

Lesson #2: Aye aye, Sir!

Artificial intelligence will become multi-modal. Future models will be able to understand not only text but also visuals, images, code, video and audio.

For digital assistants to be effective they will have to become more personalised. Finding a suitable hotel in another city, for example, is not a “right” or “wrong” task. For this, the evals (frameworks used for evaluating large language models or systems built using LLMs) will need to become much more nuanced.

The models will become smarter, become proactive (for example, read and summarise your emails) and interact as humans do.

Lesson #3: A Swiss knife on steroids

It is expected that by the year 2030 – 2032, a single system will possess about 90% of expert knowledge in every field of human activity. This will make it smarter than any single living human being. It is hard to imagine today what it will be capable of doing. There is only one absolutely clear thing – we, humans, are not ready for it.

As a final note, something to be aware of. For reasons not entirely clear only those models which have cost more than USD 100 million to train are considered a proliferation threat today. With the rapid advancement of technology, we might be facing potential dangers in future that we can’t even imagine today. Because of artificial intelligence, we all will live longer, but our lives are likely to become more dangerous.

For those willing to try and pick top lessons of their own here and here are links to the videos that are available in the public domain (not limited to subscribers only). You may also want to have a look at Daniel Kahneman’s book “Thinking, Fast and Slow” and Kai-Fu Lee’s “AI Superpowers: China, Silicon Valley, and the New World Order” (I may be receiving a commission if you decide to make a purchase). 

* from anything that you are reading, watching or hearing you can realistically expect to remember only a limited number of things. My solution is to pick just 3 items or ideas from any material. This number is non-negotiable. Even the most extraordinary experience gets compressed into 3 things to remember. This approach has worked well for me.

This note was first published on Medium on 30 November 2024.

Aivars Jurcans has more than 20 years of corporate finance and investment banking experience. His services are currently available through Murinus Advisers. More of Aivars’ writings can be found on his page Corporate Financier’s Notes.