When will AGI arrive and will it be dangerous?

Felix Naser
4 min readFeb 26

--

Much has been said about ChatGPT and how remarkably fast it has reached one million users, faster than any other online service to date: 5 days vs 2.5 months Instagram [1].

Why is it likely that the first AGI will just be a point along the continuum of intelligence [2]?

Natural intelligence vs artificial intelligence:

  • Network size: 100 billion (human brain) vs 175 billion (ChatGPT and artificial neural nets don’t have to fit into the limited brain size, but can span across supercomputer warehouses or even the whole world connected by the internet)
  • Processing frequency: 200 Hz (human brain) vs 5.8 GHz (Intel Core i9)
  • Processing speed: 1–100 m/s (human brain) vs 300,000 km/s (electricity travels along wires approximately with the speed of light)
  • Perceived dimensions: 3x space (length, width, height) + 1x time = 4 vs potentially up to 10 or more?

Artificial intelligence has the physical prerequisites to become faster and more powerful than the best natural intelligence we’re aware of. Coupling these fundamentals with recent algorithmic and performance achievements it seems likely that AGI will eventually become a reality.

Let’s look at the recent milestones and the seemingly increasing release frequency:

  1. Deep Blue (1997): IBM’s Deep Blue defeated world chess champion Garry Kasparov in a six-game match.
  2. Watson (2011): IBM’s Watson won the quiz show Jeopardy against human champions.
  3. AlexNet (2012): A deep convolutional neural network architecture that is considered a landmark in the development of deep learning and the resurgence of artificial neural networks.
  4. AlphaGo (2016): DeepMind’s AlphaGo defeated the world champion in the board game Go, marking a significant milestone in machine learning and artificial intelligence.
  5. AlphaFold (2018): AlphaFold is an artificial intelligence system developed by DeepMind that is capable of predicting the three-dimensional structure of proteins with high accuracy.
  6. MuZero (2019): MuZero is an AI system developed by DeepMind that can learn and plan in multiple domains, including chess, shogi, Go, and video games, without any prior knowledge of the game rules.
  7. DALL-E (2021): AI system developed by OpenAI that can generate novel images from textual descriptions.
  8. ChatGPT (2022): Generative Pre-trained Transformer 3 (GPT-3) is a language model developed by OpenAI and is capable of generating high-quality human-like text.
  9. LLaMA (2023): Open-source large language model released by FAIR and outperforms ChatGPT on most benchmarks [3] [4].
  10. Toolformer (2023): Language models introduced by FAIR that can teach themselves to use tools [5].

The artificial intelligence train will most likely not stop but continue to evolve further. Like other animals can’t understand how humans affect their lives, humans will most likely not be able to tell once the first AGI system has started to exist. It might happen already in our 21st century.

https://www.ted.com/talks/nick_bostrom_what_happens_when_our_computers_get_smarter_than_we_are

What can we do to increase the likelihood that AGI will have mostly positive impacts on our lives?

In order to increase the likelihood that AGI will have positive impacts on our lives, we could take following steps:

  • Prioritizing research that focuses on the ethical and social implications of AGI, and to involve a wide range of stakeholders in this process.
  • Investing in education and training programs that prepare people for the changing nature of work in a world with advanced AI systems.
  • Establishing regulations and standards that ensure the safe and responsible development and deployment of AGI systems.

In summary, the arrival of AGI is a complex and multifaceted topic that requires careful consideration. While there are risks and challenges associated with AGI, there are also fascinating opportunities for positive impact and unprecedented breakthroughs. By working together and approaching the development of AGI in a responsible and thoughtful way, we can help ensure that it benefits humanity as a whole.

--

--