What I Read Yesterday

What I Read Yesterday

• ARM, Qualcomm to make closing cases in chip license dispute trial.

• Google's Gemini is forcing contractors to rate AI responses outside their expertise

• Anti-obesity drug market

ARM, Qualcomm court case concludes this week. An adverse verdict from the jury will be bad for Qualcomm.

ARM is a UK company, listed but majority owned by Softbank. It is a semiconductor intellectual property (IP) company. The Company develops and licenses IP for various devices worldwide, and it provides development tools that accelerate product development, from sensors to smartphones to servers.

Qualcomm Incorporated operates as a multinational semiconductor and telecommunications equipment company. It has had great success in telecommunications chips which are found in Mobile Phones and it has attacked Intel and AMD by expanding into chips for computers .

The two are involved in a court case where Qualcomm has been accused of breaching a license agreement with ARM. A negative outcome in Court would be a significant setback to Qualcomm. Reuters covered the case in the article below.

(Reuters) - Lawyers for ARM Holdings and Qualcomm are set to make closing arguments on Friday in a case that could upend Qualcomm's push into the PC market with a chip meant to rival Apple and Intel on speed.

An eight-person jury in a U.S. federal court in Delaware will determine whether Qualcomm or Nuvia, a startup Qualcomm purchased for $1.4 billion in 2021, breached a license agreement with U.K.-based ARM, which supplies intellectual property to both firms.

If ARM is victorious, the British firm could force Qualcomm to destroy the technology it purchased from Nuvia, which has become the basis of a chip released this year that Microsoft and the entire Windows PC industry had hoped would claw back market share lost to Apple.

The dispute centres on the royalties that chip companies pay on each semiconductor made using ARM's technology. Before its acquisition by Qualcomm, Nuvia's license agreement required it to pay far higher royalty rates than Qualcomm.

After Qualcomm bought Nuvia, it aimed to use Nuvia's technology in its chips while paying the lower royalties due under its own agreement with ARM. ARM objected, kicking off a dispute that led ARM to sue Qualcomm in 2022.

In a trial that started Monday, ARM sought to portray Qualcomm's moves as a first-of-its-kind flouting of standard contractual terms the British company had used successfully for decades and that would have upended its business model.

Qualcomm, by contrast, portrayed ARM as a longtime vendor whose new CEO had ambitions to reap higher royalty rates and compete against Qualcomm by starting to design its own chip.

Documents and testimony also showed that San Diego, California-based Qualcomm estimated it could eventually save $1.4bn per year in payments to ARM by switching to Nuvia's technology as it pushed into new markets. Its executives repeatedly testified that their plans did not violate license agreements.

The jury must determine if Qualcomm and Nuvia breached the deals and whether ARM met its obligations under the agreements.

During the trial, jurors were shown contract language that appeared to give ARM sweeping license rights to products built using its instructions, which could have implications for ARM's agreements with other chipmakers.

Qualcomm executives testified that such a reading of the language ignored that its license covered the extensive design work done by Qualcomm engineers to create almost entirely new products.

A verdict in the trial, which started on Monday and was set to last five days, could come as early as Thursday.

Google's Gemini is forcing contractors to rate AI responses outside their expertise

Generative AI learns by its mistakes. When Chat GPT or other LLM gives an answer, it needs feedback on it so it can “learn” and get better answers next time. This process is the basic driver of improvements in AI. Models need human “Prompt Engineers” to give the feedback. The explosion in AI means the Prompt Engineers are in demand and very overworked. Are Prompt Engineers at Google being asked to do things they are not qualified to do. If so, does this spell danger about the quality and accuracy of AI models.

I read an article in Tech Crunch which suggests that this is the case.

Generative AI may look like magic, but behind the development of these systems are armies of employees at companies like Google, OpenAI, and others, known as "prompt engineers" and analysts, who rate the accuracy of chatbots’ outputs to improve their AI.

But a new internal guideline passed down from Google to contractors working on Gemini, seen by TechCrunch, has led to concerns that Gemini could be more prone to spouting out inaccurate information on highly sensitive topics, like healthcare, to regular people.

To improve Gemini, contractors working with GlobalLogic, an outsourcing firm owned by Hitachi, are routinely asked to evaluate AI-generated responses according to factors like “truthfulness."

These contractors were until recently able to “skip” certain prompts, and thus opt out of evaluating various AI-written responses to those prompts, if the prompt was way outside their domain expertise. For example, a contractor could skip a prompt that was asking a niche question about cardiology because the contractor had no scientific background.

But last week, GlobalLogic announced a change from Google that contractors are no longer allowed to skip such prompts, regardless of their own expertise.

Internal correspondence seen by TechCrunch shows that previously, the guidelines read: “If you do not have critical expertise (e.g. coding, math) to rate this prompt, please skip this task.”

But now the guidelines read: “You should not skip prompts that require specialized domain knowledge.” Instead, contractors are being told to “rate the parts of the prompt you understand” and include a note that they don’t have domain knowledge.

This has led to direct concerns about Gemini’s accuracy on certain topics, as contractors are sometimes tasked with evaluating highly technical AI responses about issues like rare diseases that they have no background in.

“I thought the point of skipping was to increase accuracy by giving it to someone better?” one contractor noted in internal correspondence, seen by TechCrunch.

Contractors can now only skip prompts in two cases: if they're "completely missing information" like the full prompt or response, or if they contain harmful content that requires special consent forms to evaluate, the new guidelines show.

Google did not respond to TechCrunch's requests for comment by press time.

Anti-obesity drug market

Morgan Stanley estimates that the market for anti-obesity drugs will grow 5X in the next six years. The obvious likely beneficiaries of this would be Novo Nordisk and Eli Lilly though there may be many others such as those who supply the Active Pharmaceutical Ingredients ( APIs) to Novo Nordisk and Eli Lilly. In addition, as the drugs go off patent, there will be opportunities for generic drug manufacturers.

20/12/2024