OpenAI reportedly warned Microsoft to move slowly on integrating GPT-4 into its Bing search engine to avoid the inaccurate and unpredictable responses it launched with. The Wall Street Journal reports that the OpenAI team warned of the negative risks of pushing out a chatbot based on an unreleased version of GPT-4 too early.

Microsoft went ahead, despite warnings that it might take time to minimize the inaccurate and strange responses. Days after Bing Chat launched in February, users discovered the chatbot was unpredictable and could insult users, lie to them, sulk, gaslight people, and even claim to identify its enemies.

Microsoft was quick to limit Bing Chat responses to stop the AI from getting real weird, and it has taken months of work to get the Bing chatbot back to a point where you can have a long back-and-forth conversation without an unexpected outburst. It still often gets things wrong, though.

https://twitter.com/tomwarren/status/1658577433954533376?ref_src=twsrc%5Etfw%7Ctwcamp%5Etweetembed%7Ctwterm%5E1658577433954533376%7Ctwgr%5E7d885fac264e9094bbabb0f9560189c2ebe1a560%7Ctwcon%5Es1_c10&ref_url=https%3A%2F%2Fwww.theverge.com%2F2023%2F6%2F13%2F23759348%2Fopenai-microsoft-bing-ai-warning-gpt-4

The Wall Street Journal’s report also describes the tensions between the two companies as they work together and compete on AI features. Microsoft executives were reportedly anxious about the launch of ChatGPT last year. The WSJ says OpenAI gave Microsoft a few weeks’ notice that it would start publicly testing ChatGPT, just as the software giant was beginning to integrate OpenAI’s models into Bing.

Microsoft and OpenAI have a rather unique partnership that has led to some conflict behind the scenes, as the two companies simultaneously support and compete with one another.

Microsoft licenses OpenAI models and technology for use across Bing, Azure, Office, Windows, and many more products. Microsoft extended this close partnership less than a month before launching its new Bing chatbot in a “multibillion dollar investment” rumored to be worth around $10 billion. Microsoft is the exclusive cloud partner for OpenAI, and Microsoft’s cloud services power all OpenAI workloads across products, API services, and research. At the same time, OpenAI has developed its own products and API services that appeal to the same customers Microsoft is courting. ChatGPT also competes with Bing AI. The Information reported on some of these troubles earlier this year.

In a Wired interview today with Microsoft CEO Satya Nadella, this competition aspect with OpenAI was briefly touched on. “I felt OpenAI was going after the same thing as us,” says Nadella. “So instead of trying to train five different foundational models, I wanted one foundation, making it a basis for a platform effect. So we partnered. They bet on us, we bet on them.”

Asked if Microsoft tried to buy OpenAI, Nadella dodged the question. I asked Bing if Microsoft has tried to acquire OpenAI, and it also didn’t want to discuss the obvious: “I could not find any information that suggests Microsoft has tried to acquire OpenAI in the past.”

In a potential landmark case for the technology sector, GitHub, Microsoft, and OpenAI have been denied their attempt to dismiss a copyright lawsuit tied to the Copilot programming assistant.

The lawsuit pertains to the AI-powered coding tool called Copilot and its underlying model Codex. Tech giants Microsoft, OpenAI, and GitHub jointly developed the said tool. It utilizes machine learning to suggest code to developers as they type.

This feature has provoked serious legal ramifications. It’s trained on publicly available code, including many open-source projects, which leads to allegations that the software may infringe on developers’ copyright.

Software developers have claimed that Copilot and Codex have been sourced from their code. Not only that, the model sometimes reproduces them without any clear permission or concern from the developers. The judge has seemingly taken the plaintiff’s claim seriously.

He has further ordered to find out if Codex’s capacity to reproduce code truly represents any software licensing term breach. Furthermore, Tigar has dodged the giants’ trial to get over a claim under section 1202 (B). It’s the digital millennium copyright act that protects the author’s or owner’s rights.

However, Tigar dismissed the claims for violating section 1202 (A) of the digital millennium copyright act. The said act covers areas like fraud, unjust enrichment, breach of the privacy policy and terms of service, violation of the California consumer privacy act, negligence, interference, and unfair competition.