Strawberry is a “thinking” neural network shrouded in mystery. OpenAI is threatened with blocking for curiosity about the new Strawberry neural network.
Social media users have started receiving warnings from OpenAI about possible blocking of access to GPT neural networks. The reason is the increased interest in the algorithms of the new version of artificial intelligence codenamed Strawberry.
We remind you that the feature of “Strawberry” is that the model pauses for 10-20 seconds to think about the request when answering. This innovation will increase the accuracy of answers and improve work with tasks requiring complex or multi-step analysis. OpenAI CTO Mira Murati called it a “new paradigm” in the development of this technology. Also, unlike current models that rely heavily on pattern recognition in their training data, OpenAI Strawberry is said to be capable of:
- Plan complex tasks ahead
- Autonomous navigation on the Internet
- Doing what OpenAI calls “deep exploration”
However, the first testers also found some drawbacks
The model, which currently supports only text format, can take too long to answer simple questions, which is not always justified. Another claimed improvement is the memorization of previous conversations for personalized answers, but this feature does not always work stably.
The development of such advanced AI tools also raises significant ethical considerations:
- Privacy concerns – How will these AI systems handle the sensitive personal data they encounter during research?
- Bias and fairness – How can we ensure that the reasoning of artificial intelligence is not influenced by biases in its training data or search results?
- Accountability – Who is responsible if an AI-driven decision causes harm?
There are still technical challenges. Ensuring the reliability and accuracy of information collected autonomously is crucial. AI also needs to be able to distinguish between reliable and unreliable sources, a task that even humans often struggle with. Furthermore, the computational resources required for such advanced reasoning capabilities are likely to be significant, raising questions about energy consumption and environmental impact.
“Strawberry” will be integrated into ChatGPT, but as a separate option. The model will also be offered under a new pricing system with limits on the number of messages per hour, and paid ChatGPT Plus users will be able to get early access.
OpenAI said it was forced to hide the details of Strawberry’s work, fearing that users would gain access to the “deep” thought processes of AI and learn about confidential information. The company does not hide its desire to maintain a competitive advantage and does not intend to disclose Strawberry secrets. Users only see the result of the neural network’s “reasoning” without explaining the decision-making process.
This policy contradicts OpenAI’s previous statements about creating open-source artificial intelligence systems with transparent principles of operation. You can get a blocking warning even for using the word “reasoning” in a request to Strawberry, regardless of the context. A paradoxical situation arises: if the customer does not understand the “train of thought” of the neural network and cannot check the correctness of its reasoning, it is difficult for him to trust the results of its work.
By the way, OpenAI is planning a significant increase in the cost of premium subscription to ChatGPT. The price may increase from $20 to $2000, although it is not yet known whether it will be a monthly or annual payment.
One of the reasons for the increase is the need for financial stability of the company, which spends about $700 thousand per day to support ChatGPT.
Conclusion
As AI models continue to evolve, we can expect to see more sophisticated applications in areas such as research, market analysis, and software development. The pursuit of artificial general intelligence is intensifying, and each breakthrough brings us closer to AI systems that can truly understand and interact with the world in ways that were previously thought impossible.