Beyond Tomorrow: OpenAI Signs Agreement With Condé Nast, Making Biocomputing The Future

Another week, another edition of Beyond Tomorrow. Let’s have a look, shall we?

For all the controversy that OpenAI has been courting these last few months, they've been making solid headway with publishers. The latest being Condé Nast. It's almost as if parent companies of news organisations have realised that the winds are changing.

In other news, we've been using silicone-based hardware for so long, not many of us are experimenting with other possible solutions. Which include and aren't relegated to squishy human brains.

On to a little little bit of housekeeping. Indian food and delivery business Zomato has acquired Paytm's ticketing business in what is largely being touted as a smart move. Good news for those looking to get into tech, startups are going to start hiring in the next few months, if a TeamLease EdTech survey is to be believed.

Finally, NDTV Profit is running a survey on young engineering graduates looking to get into AI and AI-related roles. It doesn't take more than 3 minutes to finish. Fill it in here.

Now on to the wrap!

Condé Nast Signs Agreement With OpenAI

The latest agreement leaves only a couple of the largest TV news channels in the U.S. and a few other independents. (Source:Tenor.com)

Another day, another publisher bends the knee to OpenAI. Conde Nast has joined the likes of other large publishers like Associated Press, The Atlantic, Financial Times, TIME, Vox Media, to name just a few.

The move is a bid to “expand the reach of Condé Nast’s content,” according to an internal company memo circulated by the company’s CEO Roger Lynch. Going forward, all 23 of the publisher’s content will have some sort of integration with OpenAI.

What’s more, the agreement now gives OpenAI permission to route Conde Nast’s content to both ChatGPT and its SearchGPT prototype. The AI company has said that the collaborations with its news partners will allow them to “collect feedback and insights on the design and performance of SearchGPT.”

It’s no secret at this point, that several of the world’s largest AI companies have trained their models based on existing copyrighted material. The training of their models has often come at the cost of the hard work of journalists.

In fact, in January this year, Lynch had testified in front of the U.S. Congress that AI companies’ training methods were akin to “stolen goods.” He’s also been a proponent for licensing publisher content to AI companies, something which has now materialised for Condé Nast.

Going forward, OpenAI’s SearchGPT—currently in the prototype stage—will be the one of the many tech innovations that investors, journalists, policymakers and technocrats will be keeping an eye on. It’s clear that the company is positioning the product as a competitor to Google. That’s not an easy feat mind you. But, with the U.S. Department of Justice mulling breaking up the search engine behemoth, who knows?

But I will say this: Publishers are integrations their content with AI companies. They’ve already spread to Europe. When they get to Asia, what then?

AI Compute Might Cheaper…And Fleshy

I'll take cyberpunk dystopia for $10 in another 20 years, anyone else? (Source: Tenor.com)

Artificial intelligence is data and energy hungry. A few months ago, I’d pointed out how Google, in its quest to gain a leg up in the AI race had seen its carbon footprint jump in the last few years thanks to the massive data centre requirements.

But what if there was a way to make it cheaper and more energy efficient?

Enter FinalSpark, a biotech startup in Switzerland. They’re developing biological neural networks to replace traditional silicon-based hardware in an attempt to boost AI compute power.

Research suggests that the human brain has somewhere in the ballpark of 86 billion neurons and requires 20 watts to run, that’s equivalent to a lightbulb. And all it needs is enough sleep to function optimally.

In comparison, to run the same configuration, but on silicone-based hardware would require around 10 megawatts of power, at least according to Stanford University's Kwabena Boahen. That’s a power plant’s worth of power.

It’s precisely where our current hardware struggles. Silicone-based hardware has been the main staple since at least the 1950s. Why? Mostly because we found it to be the most energy efficient while providing peak performance. That’s not changed, yet

Earlier this year, the company revealed its computing platform—called Neuroplatform—that uses organoids to power the computing. They’re basically using a cluster of lab-grown cells that are a simplified and miniaturised version of the human brain to creature computer architecture.

The way these organoids work for compute is fascinating. Much like the how neurons receive information from neurotransmitters connected to human brain, organoids are linked to electrodes which in turn are connected to computer systems.

The organoids on Neuroplatform are then basically selectively fed the feel-good juice, dopamine, as well as electric stimulation. Conceptually, its very similar to how the human brain works, just on a smaller scale. By introducing the electric stimulation and dopamine, it makes the organoids’ neurons form new pathways and connections, much like the human brain.

For now, the process is still very much trial and error, but if perfected, has the potential to revolutionise how we build AI compute, and even potentially change our dependency on silicone-based hardware.

For now, organoids on Neuroplatform can be rented by a single researcher over the Internet for $500 a month, according to FinalSpark’s website. Everybody else needs to send in applications to the company to get a personalised quote from them.