AI has been pushed for transparency in the world. Although it cannot always feel like this, most of the largest AI companies publish data regularly what they are working, they have concerns and, in the case of anthropic, in the case of anthropic, the entire report on their chatbots is completely melted.
However, Openai is taking it one step forward, recently launching its podcast. A weekly show, The Podcast delays both surface-level themes such as why they believe that the chetgt has been so popular, and a little deep things like their concerns on the future of AI.
Overall, these podcasts are the closest links that we have for internal ideas of openi – certainly the world’s largest and most powerful AI company.
Now in his second episode, what has been said so far? And is there any valuable insight that can be found from these conversations? I pigeons to bring you highlights.
Chatgpt launch
The second episode of podcast begins with a discussion, while not at all revolutionary in nature, it is quite interesting.
They discuss the launch of the chatgip, revealing some interesting points. First, the company was almost called ‘chat’, before the final-mining decision named the chat.
The chief of the chat Nick Turley states that the team thought that their matrix was broken at the launch because the device was how popular the equipment was. It went viral in Japan for three days after the launch and went viral worldwide for four days.
A day before the launch, the team was divided on whether to launch the chat or not. When 10 questions were tested before the night, it offered the acceptable answer for only half.
GPT-5 launch date
Sam Altman mentioned the launch of GPT -5 in the first episode of Podcast. So, do we finally have a launch date? No, Altman said that we are already hearing for some time that the model update will be released in ‘The Summer’.
They went to discuss the naming plans, potentially using GPT-5, GPT-5.1 etc., it will eliminate the misleading naming scheme that is seen in the past that sporadically jumps around the numbers.
While a difficult time period has been suggested for GPT-5, which may be well delayed, especially Openai has lost its team for Meta AI, Google and Deepmind researchers as well as a lot of teams.
Chat advertisement?
OpenAII has not yet introduced advertisements in all its equipment. In the first episode of Podcast, Altman emphasized the company’s will to maintain confidence, assuming that inserting advertisements in AI output may reduce credibility.
He says that other mudification options can be discovered under the line, but for now, it seems that Openai will be an advertising service.
Extremely uninterrupted chat and prejudice
There was a recent ‘smooth incident’ in Chatgpt. It noticed that the model felt highly flattery and in nature. Whereas, in theory, it sounds like a good thing, it made the model scary and more unstable in some conversations.
It was also due to excessive agreements, even when it should not be. It expressed concern about the use of the tool where pushbacks are required. For example, mental health is being tested with a chat with concerns or the decision of serious life.
He also addressed the beliefs that chatgate has become “awake” in nature, saying that neutrality is a measurement challenge, not an easy.
Mark Chen, Chief Research Officer of OpenAII, discussed this on podcast, stating that it came out of learning reinforcement from human response, inadvertently created a bias for pleasing reactions.
Chen argued that Openai quickly responded, explaining that the long -term purpose is far more important than a friendly chatbot.
He also addressed the beliefs that chatgate has become “awake” in nature, saying that neutrality is a measurement challenge, not an easy. He should focus the lapse, but should be flexible for users to interact towards their own values.
Memory capabilities
Better memory features are one of the most requested features for chatgpt. Turley predicted that, within two or three years, AI assistants would know so well to users that privacy controls and “away from records” mode would be important.
It looks like an undeniably dreadful feeling. Although it will be used, AI is able to remember important details about you with chatbots, for many people, it will feel like a large attack of privacy.
Chatgpt already has a temporary chat. It does not appear in your history and will not be added to the memory of Chatgpt or used to training objectives. Other models such as Cloud and Le Chat have created a point of being more sensitive with your data.
Turley noticed that many users are already having a relationship with AI. This, he indicates, can be both helpful and harmful. Moving forward, the team is careful with this and said that it would require careful monitoring.
Openai’s new device
Altman discussed the launch of Openai’s new device in collaboration with Jony IVE very briefly. It recently hit a huge wall when Openi prosecuted a company, claiming that he stole his views.
In the podcast, Altman said “it would be a little late” until the device comes out. He said that “the computers we use today were not designed for the world of AI.”
This, he explains, it means that they are looking for a new look on the type of technology that is more aware of your life and environment. It takes time to make something in this way, and everything else is going on in Openai, it can be a little late.
More than Tom’s guide
Back to laptop
