Content
summary Summary

According to rumors, OpenAI planned to present new products this week. However, the presentation could be canceled at short notice. Meanwhile, new model variants appear on a chatbot testing platform.

Ad

As reported by The Information, OpenAI was planning to unveil new products this week, but for unknown reasons, the presentation is being considered for cancellation. According to rumors, the company was planning to unveil a search platform as well as new AI language models such as a “GPT-4 Lite” or an update for GPT-4.

Nevertheless, there are signs of life from the new models: two new variants of the mysterious “gpt2-chatbot” model have appeared on the LMSYS chatbot testing platform: “im-a-good-gpt2-chatbot” and “im-also-a-good-gpt2-chatbot.

The name is a direct quote from a May 5 tweet by OpenAI CEO Sam Altman. According to early users, one of the models has a more recent cutoff date, meaning it has been fed more recent training data.

Ad
Ad

The performance of the models is said to be on par with GPT-4 Turbo, and in some cases even better.

Current AI systems are "laughably bad"

Meanwhile, OpenAI COO Brad Lightcap is raising expectations for GPT-5, the successor to GPT-4. When asked what we can expect from OpenAI in the next twelve months, he described current systems as "laughably bad.” Future models should be able to do complex tasks and work with you like a teammate. There will also be a move toward voice interfaces and multimodality.

“We are just scratching the surface on the full kind of set of capabilities that these systems have,” says Lightcap.

Ad
Ad
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.
Support our independent, free-access reporting. Any contribution helps and secures our future. Support now:
Bank transfer
Summary
  • According to rumors, OpenAI was planning to present new products such as a search platform and new AI language models, including "GPT-4 Lite" or an update for GPT-4, this week. However, for unknown reasons, the presentation may get canceled at short notice, reports The Information.
  • Meanwhile, two new variants of the mysterious "gpt-2-chatbot" model have appeared on the LMSYS chatbot test platform, which, according to the first users, are said to be at or above the level of GPT-4 Turbo. One of the models is said to have been fed with more recent training data.
  • OpenAI COO Brad Lightcap has high hopes for the next 12 months, describing today's systems as "laughably bad". Future models should be able to do complex work, work with you like a teammate, and offer a shift toward voice interfaces and multimodality, he said.
Max is managing editor at THE DECODER. As a trained philosopher, he deals with consciousness, AI, and the question of whether machines can really think or just pretend to.
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.