The information we have today about Project Strawberry, level 2 of the artificial intelligence race and how far we are going to go
It all started a few weeks ago with Q* nomenclature, a code name that OpenAI used to refer to what would be the next phase in its race to improve artificial intelligence and move to the next level, but what are the levels with which OpenAI imagines the future?
Well, it was OpenAI itself that determined the roadmap of artificial intelligence in 5 levels or phases:
Level 1: Text Bots
This is the moment in which we find ourselves today. We have Chatbots like ChatGPT, Claude or CALL that are able to respond and understand what we say and give us a response in accordance with the learning that these language models have had.
I'm sure you've tried some. You have to give them a prompt —Check out the article on how to write the best— which essentially means that they require instructions in order to give us an answer, whether it be writing a text, giving us information, etc.
At this first level - which although it is the first has been revolutionary - We have discovered the potential of these artificial intelligences, an absolutely amazing experience game changer within the world of technology but which is currently "limited" to its own learning and the data it has.
This limitation does not prevent language models from giving us original answers, but they are always based on what they already know. They do not reason, and it is here that we move to level 2.
Level 2: The “reasoners”
Reaching reasoners would be an important milestone in the development of artificial intelligence as it promises to transform the way we interact with technology. This phase focuses on overcoming the current limitations of AI, such as a lack of complex reasoning and the inability to perform tasks that require long-term planning and decision-making.
We would be talking about being able to make decisions at a human level, being able to interpret the instructions they receive and give answers based on that reasoning.
The Phase 2 is distinguished by its focus on improving the reasoning of AI models. Current systems, such as chatbots, are capable of generating responses based on trained data, but lack the ability to reason effectively and perform complex tasks without human intervention.
This opens the door to a future in which language models are not only capable of giving us more or less simple answers, but We will be able to observe to what extent they are able to reason like people.
OpenAI seeks to solve these limitations through innovations that will allow AI to not only generate text, but also understand context and plan ahead. Can you imagine?
This approach not only has implications for improving AI functionality in everyday applications, but also opens up new possibilities in fields such as scientific research, where AI's ability to reason and plan could be a paradigm shift.
Well it seems that the Project Strawberry would be the name to define this jump from Level 1 to Level 2 and on which OpenAI is working. For now we are going to skip the rest of the levels and focus on the most immediate.
Project Strawberry: Boosting Reasoning in AI
At the heart of OpenAI Phase 2 is Project Strawberry, an initiative purportedly designed to take AI’s reasoning capabilities to the next level.
Previously known as “Q*,” Strawberry is not simply a new AI model, but an innovative technology that seeks to improve the performance of existing models through an advanced post-training approach.
The @iruletheworldmo X account has been interacting with accounts such as that of Sam Altman himself, CEO of OpenAI, exchanging some messages that directed attention to this Strawberry Project in a surreptitious manner…
The goal of Strawberry is to enable AI models to not only answer queries, but They are also able to reason and perhaps perform complex, long-term tasks that require planning., execution of a series of actions and the ability to browse the Internet autonomously to carry out in-depth research.
This progress is essential to overcome the Current barriers to AI, which often falls short on tasks that require a higher level of understanding and reasoning.
Inspired by techniques from institutions like Stanford, this approach allows AI to Refine your intelligence after initial training, addressing problems that previously seemed insurmountable for AI systems, such as lack of common sense.
Although specific details about how Project Strawberry works are being kept under wraps, what is known so far suggests that It could be a revolutionary change in the way we understand and we use artificial intelligence.
This project has the potential to take AI beyond simple generated responses, towards true reasoning and planning capabilities, bringing it one step closer to general artificial intelligence.
Give free rein to your imagination
I asked ChatGPT what it would mean for humanity if Artificial Intelligence reached that level of human reasoning. Attentive to your reply:
Can you imagine a world like this?
