Better than humans? AI is moving towards AGI

Breakthroughs in artificial intelligence (AI) are accelerating. AI technology is already being used for a myriad of applications, from tackling climate change to space exploration, develop cancer therapies and providing real-world navigation for robots. The number of studies paper focused on AI has grown so fast in recent years that it almost seems exponential. While we’re still a long way from widespread adoption of AI across all spheres of human endeavor, it’s safe to say that the technology has now crossed the divide between early adopters of new and little-known products and mass adoption by mainstream users.

The most talked-about AI breakthrough of the year is the new category of generative AIon which is based big language models. Almost overnight, a proliferation of image generation tools appeared, including OpenAI’s DALL-E, Google’s Imagen, Stability.ai’s Stable Diffusion, and Midjourney. I wrote a few months ago about the disruptive impact of these tools on creative professions ranging from digital artists to programmers.

As far-reaching as these developments are, the new conversational text bot may be more important ChatGPT, also from OpenAI and based on GPT-3.5. It has been trained on a huge amount of text data from various online sources. It can chat, answer questions, create plays and articles, write and debug code, take tests, manipulate data, give advice and tutoring, among other things.

ChatGPT has been discussed a lot online, including by savvy reporters Kevin Roose in the New York Time and Derek Thompson at the Atlantic Ocean. Thompson calls this and other recent generative AI tools a “second mind for the creative class.” Roose wrote that there is already a comparison with ChatGPT the iPhone in terms of the potential impact on society.”

However, ChatGPT is still in its infancy. Nearly everyone, including OpenAI, recognizes that the technology is far from a perfected product, as evidenced by the “occasionally incorrect information‘It generates. Nevertheless, as Jack Clark states in his Import AI newsletter: “In a few years these systems can be better than humans, which will have wild implications.”

AI: new master of strategy

While these remarkable generative AI highlights are hugely important, several other recent AI advancements could ultimately have an even bigger impact on the world. An example is the recent AI defeat human experts in Stratego, a two-player strategy war game that requires long-term thinking, bluffing, and strategy. Deep Mind’s DeepNash algorithm, a trained autonomous agent that can develop human-level expertise, powers the AI ​​that plays Stratego. DeepNash is based on an entirely new approach to algorithms using game theory and model-free learning with deep reinforcement.

Unlike chess and Go, Stratego is a game of imperfect information: players cannot directly perceive the identity of their opponent’s pieces. It is considered one of the most difficult games, due to the seemingly infinite number of possible moves (as many as 10⁵³⁵), more than even the notoriously complex Go (10³⁶⁰). To win, DeepNash combined both long-term strategy and short-term decisions, such as bluffing and risk-taking, a unique ability for an AI.

As reported by peculiarity Centerthe researchers stated, “By creating a generalizable AI system that is robust in the face of uncertainty, we hope to push AI’s problem-solving capabilities further into our inherently unpredictable world.”

The art of diplomacy

Speaking of unpredictability, Meta recently unveiled “Cicero” – an AI system named after the classical statesman and scholar who witnessed the fall of the Roman Republic – that defeated humans in another strategic war game, diplomacy.

Unlike Stratego, Chess or Go – all of which are zero-sum, winner-takes-all competitions – diplomacy is collaborative and competitive at the same time. Up to seven players compete, negotiating using deceit and cooperation, trust and betrayal, to form and break alliances in the pursuit of total domination. In other words, diplomacy is much like real-life strategic negotiations between multiple competing entities, be they gamers, companies, or countries. As reported by Gizmodo“to ‘win’ at diplomacy [the AI] must understand both the rules of the game efficiently [and] fundamentally understand human interactions, deceptions and cooperation.”

This rich opportunity gets to the heart of what meta wanted to develop: “Can we build more effective and flexible agents that can use language to negotiate, persuade, and collaborate with people to achieve strategic goals, similar to the way humans do?” The company claims that Cicero achieved more than double the average score of the people who played it webDiplomacy.net and ranked in the top 10% of participants who played more than one game.

Meta positions Cicero as a research breakthrough that combines two distinct areas of AI: strategic reasoning and natural language processing. According to to three-time world diplomacy champion Andrew Goff: “Cicero is resilient, relentless and patient.” He adds, “It makes the best decision not only for itself, but for the people it works with.”

Generalized AI

“Narrow AI” contains algorithms that do only one thing, albeit extremely well, such as recommending which book you might like based on books you’ve previously viewed on an ecommerce site. A narrow AI algorithm cannot effectively transfer nothing it has learned to another algorithm designed to fulfill a different specific purpose.

The other end of the AI ​​spectrum is considered “strong AI” or alternatively artificial general intelligence (AGI). Probably every AI expert will agree that this does not exist today and remains in the realm of science fiction. If and when AGI is achieved, it would be a single AI system – or possibly a group of linked systems – that could be applied to any task or problem because it could act and think like humans.

Murray Shanahan, a professor of cognitive robotics at Imperial College London, said of the exponential representation podcast that AGI is “in a way as smart as humans, and capable of the same level of generalization that humans are capable of, and possesses the common sense that humans have.” This is very similar to the capabilities of this new wave of strategy algorithms.

However, there is no single AGI definition. For example, Elon Musk doesn’t think ChatGPT qualifies because it hasn’t invented anything great:

Towards Artificial General Intelligence (AGI)

At least by these criteria, ChatGPT is not an AGI, nor is DeepNash or Cicero. However, what they all have in common is clear progress in this direction. As Stuart Russell, a professor of computer science at the University of California and a leading artificial intelligence researcher, points out, “You won’t be able to pinpoint the actual date of arrival of AI for general use; it’s not just any day. It’s not like it’s all or nothing. The consequences [of AI] will increase. So with every advancement of AI, it expands the range of tasks significantly.”

With each passing year, we can expect to see much greater capabilities heading into AGI as these models become more sophisticated and new systems emerge.

Given the pace and scope of recent AI breakthroughs and the massive growth in the number of research papers, we can expect developments to accelerate, with profound implications for work and life. For example, within a few years, ChatGPT or a similar system could become an app similar to Samantha in the movie Hair. ChatGPT is already doing some of what Samantha did: an AI that remembers past conversations, develops insights based on those conversations, provides helpful guidance, and therapy and can do so simultaneously with thousands of users. Or imagine NATO using tools like DeepNash or Cicero with its members or in negotiations with rivals.

We are witnessing growing momentum towards AGI, though experts estimation of its appearance time is 2045. AGI or not, AI technology is becoming more advanced and deeply embedded in the fabric of our lives.

Gary Grossman is senior VP of Technology Practice at Nobleman and global leader of the Edelman AI Center of Excellence.