OpenAi found itself in the news this week as some took to X and Reddit to describe how ChatGPT started unprompted conversations. This set off much debate about whether the posters were faking the story (and screenshots) or it was real.
Evidently it happened although OpenAI asserted it was a bug in the software. The company claimed it fixed the issue that was causing messaging.
Do you believe this? Was this simply a bug or was Altman and company floating it out there to see the response?
We cannot answer that. It does, however, tells us a great deal about data and what these companies are going to be doing.
Image generated by Ideogram
Unprompted Chat And The Thirst For Data
Even if this is removed from the equation, it is only temporary. We are going to see more unprompted interactions going forward.
Why do I make this claim?
When it comes to data, this serves two purposes.
It starts with the need for more. If a ChatGPT, for example, starts 100 million conversations, how many get responded to? Even if the number is 1%, that is 1 million conversations. Some will go in depth, providing a vast amount of data.
This is something that could be prompted each day. Over time, as people get use to it, chatting with the machine will seem normal. It is interesting how there is a percentage of the population that cannot stop talking about themselves. If prompted to do that, they will succumb even if they are simply interacting with software.
Over time, we can see how this keeps increasing the amount of data an OpenAI will have access to. For this company, there is incentive since they do not have a social media platform to garner data each day. This is an advantage that X, Google, and Meta have.
The second reason why this is of benefit to the companies is because of the human-synthetic connection.
More of the data online is going to be synthetic. This is generated each time someone does a prompt. With the increase in utility, the output by the machines will only accelerate.
Many are questioning the value of this data, especially as it is cycled through repeatedly. There is a contingency that thinks degradation is the only result. Perhaps this is the case.
If so, human interaction with the data is the solution. As more synthetic data is pushed out, human feedback helps the knowledge graph. The data is no longer purely synthetic. Instead, more context is provided, something that is essential if these models want to contain knowledge as opposed to simply information.
To me, these are two enormous benefits for a company like OpenAI (or any of them really).
Web 3.0 Entry
The problem with the present race in AI (actually language models) is the fact that we are looking at closed entities. Even something like Meta, which open sourced Llama, keeps the data separate. Opening the weights is a step in the right direction yet it is not truly open.
Without the data, a team that is developing something is not going very far. It is still dependent upon Meta to provide the updates with the new data used for training.
Web 3.0 could be a major boom to what was described above. The idea of conversations on-chain is nothing new. As these models improve, people can be openly "conversing" regardless of the source. For now, it is still possible to notice a difference between humans and machines.
Of course, all of this gets taken to another level when voice conversations are involved. OpenAI turned a lot of heads with its conversation piece when it released GPT4.o. That feature is still unreleased. Others are creating open source versions of this, elevating the entire industry.
Conversations are recorded. Anyone who interacts with Siri or Alexa is providing larges amounts of data to these companies. This will only expand as more people start to utilize the technology.
As always, this comes back to where the data is stored and who has access? With OpenAI and the likes, we know the answer. It is all theirs.
Even if you believe the unprompted chats was a bug, the fact that it is able to go through older conversations and ask questions shows the progress that is being made. It also highlights who owns the data.
Anything you do with ChatGPT is there, data that is closed off to the rest of the world. The company is also going to utilize it however it sees fit.
Watching the actions by these companies shows where the value is and what is needed. Web 3.0 should learn these lessons to counter what Big Tech is doing.
It is all there before us.