Scary or impressive: either way. That’s how many see the evolution of artificial intelligence (AI), which will become more advanced. Apparently, a new GPT is on the way that will become more human and less prone to error than we already know.
As mentioned Robot bodyAccording to a Reuters report, OpenAI has developed “Strawberry.” Essentially, this will be a more complete and detailed version of what Chat GPT already is.
Continuous Improvement ‘May Threaten Humanity’ – OpenAI Employee Says
The situation is new and the evidence for this is the testimony provided by OpenAI employees. According to the same source, there are already those who say that the new platform “could threaten humanity,” whatever that means, in the context in question.
Whether or not this is the case remains to be seen, especially since humans are already known to be “animals of habit.” However, what we do know at this point is that Strawberries are being trained to do very deep research. This essentially means that typical GPT chatbot accuracy errors no longer occur.
As Bloomberg puts it, this is expected to assume “human-like thinking,” from the way it searches, to the autonomy with which it does so. Right now, we’re only getting a glimpse of the superficial data, but we already know that Strawberry will be able to learn from its mistakes.
AI must be able to learn from its mistakes.
In practice, this method should work based on the Self-Taught Thinker (STAR) method. This means that the AI can learn, on its own, and in a super-efficient way, what is wrong and what is right, as well as correct its own thinking.
As for these new tools, it is estimated that they will be able to work autonomously. What makes us confused is the word “autonomous”. After all, what does it mean in this case?
Does it look for topics that interest it itself? Does it automatically correct errors, without you telling it when it failed? At this point, there is a lot of uncertainty in this regard, but when we hear employees themselves give testimonies like the one mentioned above, we can’t help but be curious.
It is not known when “Strawberry” will definitely debut.
Also expressing his concern was a Stanford professor. “I think it’s exciting and scary at the same time… If things continue to go in this direction, we have some serious things to think about as humans,” says Noah Goodman (via Reuters).
At this time, we also do not know whether this improved “strawberry” logic will debut with GPT-4o or with GPT-5. There is also no specific release date for this feature, which is currently under development.
