AGI is the next natural step for AI. Is AGI less than 20 years away?
AI development is rapid
In the last year and a half, AI has become "hot." Most companies and individuals have recognized the potential of language models like ChatGPT to enhance productivity and quality. At the same time, users' questions have provided valuable feedback to improve the algorithms.
Among developers, there's a real race, with Microsoft's investment in OpenAI already reaching $13 billion. OpenAI's Sam Altman has been in discussions with several Sovereign Wealth Funds since early spring for a seed round to develop and manufacture chips and quantum computers for up to 7 trillion USD! This is the world's largest seed round ever.
Next Milestone: Integrating with AI
The next significant AI milestone will be when each of us integrates AI into our daily decisions. In other words, our priorities, judgment, and ethics. When we involve AI here, it truly learns about us.
- All the major tech companies are currently focused on products that can function as "personal assistants," as both Satya Nadella and Sundar Pichai have expressed it.
- Apple’s new major AI focus, after abandoning its EV project in February, is also in the area of "personal products."
- The tech companies collectively expect that we will converge with AI, making it a part of ourselves.
However, this requires both willingness from consumers and capability from producers.
Consumers are willing ...
Consumers are very willing. For example, the second most used AI service at the beginning of the year was the chatbot Character.ai, which can help with learning new languages or brainstorming ideas. But it can also emulate chats with specific people, such as Elvis Presley, Michael Jackson, or a psychologist. The developers of the chatbot have noted that users are very willing to open up to machines, sharing confidential thoughts and dilemmas more than they would typically share with their partners or healthcare professionals.
This is why the commercial AI opportunities for "personal services" are so enormous. This also leads to predictive capabilities.
... and hence AIs Capability Is Coming Soon
The capability is currently the challenging part of the equation because AI needs to be better at judging ethics than the individual. AI will only be relevant if it improves our chances of making good choices.
Ethics applies to everyone, and therefore, it can be described in syntax, such as legal text. AI algorithms can work with this. What is still needed is learning from practice and computational power.
Learning from individuals is necessary because ethics is, for example, relative to the observer's perspective and the situation, even legally. Therefore, AI needs to learn to understand the framework for this. Next, it needs to learn how ethics change over time and that ethics is mixed with certain absolute ideals like the UN's Sustainable Development Goals. The more users AI can follow through, for example, chatbots, the better it learns.
Computational power depends on physical hardware and how effectively the algorithm develops itself. The hardware limitations today need to be solved, among other things, through Sam Altman's seed round mentioned above.
Maybe Even Sooner Than We Thought
Self-learning capability is surprisingly high. In 2021, OpenAI started a project called Q*, to develop AGI with, among other things, consciousness and intentionality (Deep AI). In November 2023, parts of the project had to be "scaled back and focus adjustedprimarily because the learning speed of the algorithm exceeded anything seen before or calculated as possible. In other words, it was an early sign of AGI. The success scared OpenAI's board, and Sam Altman was unemployed for a couple of days.
Since then, the project has continued behind closed doors due to the lawsuits Elon Musk has (re)filed against OpenAI. However, Q* is not alone in the lead. Google's DeepMind is rumored to be just as advanced in AGI development.
Just over a year ago, most AI professionals estimated that AGI was about 20 years away.
But if the industry's new expectations are met, that time frame could be just a few years out..
The speed of self-learning is now growing exponentially in almost all areas.
Are we ready for that? Can anyone "tame" AGI, and if so, who will control such an enormous power potential? Will we even notice that AGI exists if it is smarter than us?