A step closer to AGI

Welcome to the 9 new deep divers who joined us since last Wednesday. If you haven’t already, subscribe and join our community in receiving weekly AI insights, updates, and interviews with industry experts straight to your feed

DeepDive

Your weekly immersion in AI. 

On 11th July 2024, OpenAI shared its new AI classification system with all its employees – laying out five levels to track its progress towards creating AI software that has the capacity to outperform human intelligence. 

What are the levels?

As reported by Bloomberg, the five levels are as follows:

Level 1: Chatbots – AI that can engage in conversational language 

Level 2: Reasoners – AI with human-level problem solving

Level 3: Agents – AI systems that can take actions

Level 4: Innovators – AI that could come up with new inventions

Level 5:  Organisations – AI that could do all the work of an entire organisation

We’re already at level 1. OpenAI claims to be very close to level 2 at this point in time; and then at level 3, OpenAI suggests that ‘agents’ could spend a number of days taking actions on behalf of a user, without requiring user input. 

Working on level 2

OpenAI predicts that we’ll move through each of these levels over time. Shortly after releasing this classification system, Reuters reported on an internal OpenAI document it had seen in May; which details the development of a new reasoning technology, currently referred to by the code name ‘Strawberry’. 

Strawberry is a ‘work in progress’ with no clear release date, and we don’t know much about it yet – with Reuter’s source noting that its workings are kept in secrecy, even within OpenAI. 

But the document describes a project that uses Strawberry to enable AI models to plan ahead, autonomously navigate the internet, and perform ‘deep research’ on users’ behalf.

Reuters asked OpenAI for a statement on Strawberry, and spokesperson responded:

“We want our AI models to see and understand the world more like we do. Continuous research into new AI capabilities is a common practice in the industry, with a shared belief that these systems will improve in reasoning over time.”

How close are we to AGI? 

Predictions remain varied. 

• In a 2022 Expert Survey on Progress in AI, 50% of respondents said they believe high-level machine intelligence could be operational by 2059.
• Shane Legg (Co-Founder of Google’s DeepMind) said in an interview with tech podcaster Dwarkesh Patel that there’s a 50% chance AGI could be achieved by 2028.
• Before the release of the new classification levels, Sam Altman (CEO at OpenAI) said he expects AGI to exist in the “reasonably close-ish future.”
• And Elon Musk predicts that full AGI will be in existence by 2029. 

We want to know what you think. How close are we to AGI – and more importantly, what does this mean for the relationship between humans and technology? 

Open this newsletter on LinkedIn and tell us in the comment section. We might reach out to you to feature your perspective in a future newsletter. 

Immerse yourself in the AI ecosystem at DeepFest 2025

Registrations are open now. Meet us in Riyadh in February 2025 to meet the leading minds in AI research and development, and position yourself for success in an AI-powered business ecosystem.