For about two months, we dedicated Fridays to discussing AGI in our posts. We explored different approaches, methods, and systems that could potentially help achieve human-level reasoning and enhance overall AI systems' intelligence. Even now, nobody knows exactly which technology will be the key to AGI, and it’s interesting to analyze past and ongoing research to find what could work for more advanced AI. We’ll continue to investigate this area, but now we would like to summarize what we have already explored.
Here is our collection of 8+ researches that could be used to achieve (maybe!) AGI:
Three Google DeepMind’s milestones:
Alpha Go, created to perfectly play the Go game, uses two neural networks combined with reinforcement learning to enhance AI reasoning and creativity. → Read more
Alpha Fold, a tool for accurate protein interaction predictions, implements attention mechanisms to understand relationships of amino acid sequence and graph neural networks to represent proteins as networks. → Read more
AlphaProof and AlphaGeometry 2 are created for advanced math reasoning. These legendary systems got silver medal at the International Math Olympiad thanks to the following approaches:
AlphaGeometry merges a neural language model for pattern recognition with a symbolic deduction engine for decision-making.
AlphaProf uses AlphaGo's reinforcement algorithm for math reasoning and translates natural problems into Lean, generating new math problems.
A path towards AGI: 3 @GoogleDeepMind's milestones on the way to state-of-the-art AI systems.
Hoping to achieve AGI, many from ML community focuses on LLMs.
But there are other ways to enhance AI systems' intelligence, and achieve (maybe!) human-level reasoning.From now on, we… x.com/i/web/status/1…
— TuringPost (@TheTuringPost)
11:12 PM • Jul 26, 2024
Neuro-symbolic AI systems are hybrid architectures combining neural networks, which excel at pattern recognition and intuitive ideas, with symbolic reasoning methods that emphasize logic, rules, and structured knowledge. They mimic the human use of both logic and intuition in decision-making. → Read more
A path towards AGI: Neuro-symbolic systems
Sometimes when you can't find the solution it's useful to look back and reflect on past approaches.
Since 2015, various neuro-symbolic systems have emerged, including:
- IBM's researches
- MIT's Neuro-Symbolic Concept Learner (NS-CL)
-… x.com/i/web/status/1…— TuringPost (@TheTuringPost)
12:27 AM • Aug 3, 2024
The LLM-empowered Autonomous Agent (LAA) demonstrates the synergy of symbolic and connectionist AI. It uses a neuro-vector-symbolic approach, combining neural networks, vector representations, and symbolic reasoning. Vector part of the system helps to handle large datasets and perform tasks like in-context learning. LAA is an example of more autonomous and intelligent system, that can handle a wide range of tasks. → Read more
A path towards AGI: neuro-vector-symbolic approach
Earlier we wrote about neuro-symbolic systems as an approach that mimics human use of both logic and intuition in decision-making. And now, new interesting research based on neuro-symbolic AI was released: “Converging Paradigms:… x.com/i/web/status/1…
— TuringPost (@TheTuringPost)
1:49 PM • Aug 9, 2024
Machine Psychology introduces four paradigms to understand LLM behaviors: heuristics and biases, social interactions, psychology of language, and learning. Researchers aim to use human psychology-inspired experiments to explore LLMs' deeper workings. → Read more
A path towards AGI: Machine Psychology
Researchers from the University of Stuttgart, Google DeepMind, Helmholtz Institute, and TU Munich are pioneering a new field dubbed "machine psychology." By applying behavioral experiments inspired by human psychology, they aim to uncover… x.com/i/web/status/1…
— TuringPost (@TheTuringPost)
10:14 PM • Aug 16, 2024
Self-Play Mutual Reasoning (rStar) involves two models discussing a problem to find a solution. The first model generates reasoning paths using Monte Carlo Tree Search (MCTS), and the second model verifies them. MCTS employs a rich set of human-like reasoning actions to offer multiple high-quality answers. → Read more
For a deep dive into how knowledge-grounded reasoning compares to chain-of-thought, see our Chain-of-Knowledge explainer
A path towards AGI: Self-Play Mutual Reasoning (rStar)
What if language models could discuss a problem to arrive at a better solution together, like people usually do?
Last week, researchers from Microsoft Research Asia and Harvard University introduced rStar, a self-play… x.com/i/web/status/1…
— TuringPost (@TheTuringPost)
11:29 PM • Aug 23, 2024
STRATEGIST is a method designed to help LLMs learn and improve strategic skills, especially in multi-agent games where they need to "outthink" other players. It uses Monte Carlo Tree Search (MCTS) to improve decisions, teaches the model to "think" on bi-levels (high-level strategy and low-level actions) and generates dialogues in games. → Read more
A path towards AGI: Strategic skills, games and self-training for complex decision-making
What if we could achieve AGI by teaching systems general intelligence through games? It could be similar to how people learn different life scenarios in childhood while playing games.
We… x.com/i/web/status/1…
— TuringPost (@TheTuringPost)
11:35 PM • Aug 30, 2024
Here are 4 studies that discuss how good are LLMs at creating novel research ideas:
Can LLMs Generate Novel Research Ideas? A Large-Scale Human Study with 100+ NLP Researchers → Read more
SCIMON : Scientific Inspiration Machines Optimized for Novelty → Read more
Can Large Language Models Unlock Novel Scientific Research Ideas? → Read more
The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery → Read more
A path towards AGI: LLMs that generate novel ideas
What if we could enhance LLMs so that they can generate novel ideas independently and go further in their own decision-making?
Several studies analyzed LLMs' creativity in generating new research ideas, showing the tendency of… x.com/i/web/status/1…
— TuringPost (@TheTuringPost)
11:00 PM • Sep 13, 2024
In a survey on attention heads, researchers compared LLMs’ function to human brain processes through four steps: Knowledge Recalling, In-Context Identification, Latent Reasoning, and Expression Preparation. They also propose to apply psychological concepts to AI to enhance human-like thinking and behavior in machines. → Read more
A path towards AGI: Attention heads
What if attention heads in transformers worked more like human brains?
In a recent survey, researchers from IAAR, AIR, and Tsinghua University sought to understand how attention heads function by comparing them to how the human brain thinks.… x.com/i/web/status/1…
— TuringPost (@TheTuringPost)
11:00 PM • Sep 20, 2024
📌 Practitioners on AGI: For builder perspectives on what AGI means in practice, see our interviews with Anneka Gupta (Rubrik) on agentic AI safety and Ben Goodger (OpenAI) on the browser as AGI's first body.
📌For a deeper look at why defining AGI is so elusive, see The Elusive Definition of AGI by Charles Fadel.







