Beyond Tomorrow: Whistleblower Protections, AI Boosting Top Line For IT — Weekly AI Roundup

Good news! We have OpenAI adjacent news for the week. It seems folks at the artificial intelligence company decided to stop hogging the limelight and share it with others.

In last week's newsletter, we made mention of research by folks at the University of Chicago's Booth School of Business using Chat GPT-4 to do financial analysis. We did a full breakdown of that paper and its findings earlier this week. You can read it here.

Now, on to this week's wrap!

Artificial intelligence companies have been around for a while now. That means that companies are trying to find a good way to monetise the technology, which is reasonable enough. But the way AI and GenAI is evolving, there have to be certain considerations made when it comes to prioritising safety and control of these systems. That's why whistleblowers exist.

But they don't exist in a vaccum, especially considering there are usually special considerations for such employees in other industries.

In an open letter published online, 11 OpenAI and two Google DeepMind former and current employees have called for wider protections for whistleblowers.

Explaining their rationale for the open letter, the group wrote that in the current landscape, AI companies have "strong financial incentives to avoid effective oversight" and that they "don't believe that bespoke structures of corporate governance are sufficient".

They're right, of course. Currently, regulation surrounding AI is vague, esoteric and not enforceable, unless you're counting the European Union. Here in India, we only have suggested guidelines, while in the US they're still mulling laws around regulating AI.

Since there is "no effective government oversight of these corporations", only current and former employees are really the few people who can hold them accountable, the group said in its letter.

The devil's dozen mentioned four principles in their letter, asking that AI companies adhere to them. Here's what they said:

1) Companies won't enter into agreements with employees that prohibit them from voicing criticism of the company related to concerns about risk.

2) AI companies will create anonymised processes for employees to submit risk-related concerns to senior management.

3) Companies will create a supportive culture of open criticism which allows both former and current employees to raise concerns.

4) Companies won't retaliate against former or current employees for sharing risk-related confidential information publicly.

A team of researchers at Microsoft Inc. have developed an AI foundation model, Aurora, that has been trained on over 1 million hours of atmospheric data to help accurately provide weather forecast and give insights.

Foundation models, in case you were wondering, are large deep learning neural networks (a type of machine learning that simulate the complex decision-making power of a human brain), that serve as the base on which other machine learning or Gen AI models are built.

As a result of Aurora's training, the 1.3 billion parameter model is able to excel at prediction task in both regions, which don't have a lot of data and extreme-weather scenarios. The model is capable of producing five-day global air pollution predictions as well as a 10-day high-resolution weather forecast, all in under 60 seconds.

The really cool part, though, is the fact that Aurora can basically do more, the more diverse data sets you feed to it. More diversity means Aurora is able to learn a "more robust and generalisable representation of atmospheric dynamics."

Is the weatherman on your local TV channel finally going to have competition? Looks that way.

Intel and AMD have had enough of Nvidia's dominance in the AI chip market and both seem to desperately want a piece of that pie. It remains to be seen if they'll be able to make a dent in the latter's iron grip on the industry.

From June 4 to 7, the three behemoths of the chip industry battled it out on the presentation stage of the Computex 2024 conference in Taiwan's capital city of Taipei.

Here's a quick breakdown of what each of them showcased:

Intel: The American tech company announced three things at Computex. First was Lunar Lake, an upcoming processor ecosystem with a focus on improved efficiency and better performance for laptops. Second was the Intel Xeon 6 family of processors, which is for a variety of use cases and workloads including AI, high-performance computing, image processing and data analytics. Lastly, Intel Gaudi, a generative AI kit with two Intel Gaudi 2 accelerators at $65,000 dollars, which the company claims is one-third the cost of comparable competitive platforms.

AMD: Set to go on sale from the fourth quarter of fiscal 2025, the company's new Instinct MI325X chip is for powering end-to-end AI infrastructure from the data center to PCs. Additionally, they also showed off their latest Ryzen systems which will be going into Microsoft's Copilot+PCs.

Nvidia Corp.: Among the host of announcement made by the company, the largest was the reveal of their new Rubin platform, which is set to succeed upcoming Blackwell. The new platform will feature new graphic processing units, a new ARM computer architecture-based CPU, Vera, and a host of advanced networking capabilities with other Nvidia tech.

India's IT companies are optimistic about the revenue potential of Generative AI, but experts remain cautious.

"Yes, it is definite that going forward every company will want to adopt AI and see how they can use AI in their organisation, but net-to-net, will it cannibalise into something else, we don't know," said Hiren Ved, director and chief investment officer at Alchemy Capital Management.

However, given AI's potential to improve efficiency and reduce costs, there's no reason for people to see it as anything but a positive long-term trend.

You can read the full story here.

Beyond Tomorrow is a weekly newsletter sent to your inbox every Saturday to give you a roundup of everything AI in the last week.