KATE THOMPSON DAVY: New leaps and bounds for the AI poster-problem child

The machine knows, and now it speaks! It hears you, and it responds. Or, to be less dramatic, the artificial intelligence (AI) powered chatbot we know as OpenAI’s ChatGPT can officially hear and respond to a user, audibly.

This is the latest tweak of this incredibly popular tool, which OpenAI says will be available to paying customers within the next two weeks, and more broadly to nonpaying users “soon after”.

The updates all focus on how one uses the ChatGPT interface. The new version will let you upload images or speak to the AI — as you might already do with Siri on iPhone or Amazon’s Alexa — by tapping a button before you talk.

It will be interesting to see what this does to the lightning quick response time we know and love about ChatGPT. The model uses a text-to-speech system that will convert your spoken word to text and then run that through its large language model to extract a response, which is then converted back to speech and vocalised by the interface.

Verge reports: “It should feel just like talking to Alexa or Google Assistant, only — OpenAI hopes — the answers will be better thanks to the improved underlying tech. It appears most virtual assistants are being rebuilt to rely on [large language models] — OpenAI is just ahead of the game.”

OpenAI’s announcement gives a handful of details about how this might work, and the use cases it foresees, saying: “You can now use voice to engage in a back-and-forth conversation with your assistant. Speak with it on the go, request a bedtime story for your family, or settle a dinner table debate.”

It makes sense to me that OpenAI is investing in new interface methods, but what effect that might have on the well-established incumbents I am not sure. I suspect we will see divergent uses, people sticking with Siri (for example) for one kind of request, and using ChatGPT for others. Maybe there will be a catastrophic (for Siri et al) eclipsing, but that would require overcoming user inertia — a remarkably sticky element.

Either way, OpenAI continues to occupy the role of “thorn in our side” for all other tech firms in 2023. OpenAI might also be hoping that these kinds of updates will drive another wave of eager users and even funders.

Not to be left out, Amazon announced on Monday that it would be investing up to $4bn — starting with $1.25bn immediately — into Anthropic, a start-up that describes itself as “an AI safety and research company based in San Francisco” with a clear focus on ethical AI principles and practices, emphasising trust and research across its website.

Reuters calls this deal an “effort to compete with growing cloud rivals on artificial intelligence”, the largest of those cloud rivals being Microsoft and Google, who is also the Anthropic funder. One of its founders was also a former OpenAI executive, just to keep it all in the family. Amazon cloud customers and employees will get early access to Anthropic technology, which includes an AI-drive assistant called Claude.

Generative AI is a field seeing immense investment and interest across the world, as anyone with a vague awareness of the news this past year will surely know by now. We have been bombarded with AI news and updates since ChatGPT put generative AI on everyone’s radar in November 2022. I’ve written a fair whack of it myself.

Still, the hype is not going away any time soon. Even as much of the news about AI has shifted to the matter of regulation and ethical standards, the leaps and bounds continue — driven by a hunger for the productivity boost economists and analysts expect this tech to drive.

One number we see floating to the surface quite often comes from Accenture, which in 2017 — that’s pre-ChatGPT — claimed AI would boost average profitability by 38% and through that add about $14-trillion to the economy. More recently, earlier this year PwC researchers suggested global GDP might be 14% higher in 2030 due to AI. That’s about $15.7-trillion, which PwC calls “the biggest commercial opportunity in today’s fast-changing economy”.

But — remixing what author William Gibson presciently wrote — that future won’t be evenly distributed when it gets here. PwC expects China to see the greatest gains as a result of AI, up to 26% of GDP in 2030. North America comes in second with a “potential 14%” — quite a gap between first and second. Industries expected to see the largest gains include financial services, retail and healthcare.

As MIT’s The Algorithm newsletter pointed out in the latest edition, it has been six months since the Future of Life Institute (FLI) published its now infamous open letter calling for a moratorium on the development of powerful generative AI tools akin to ChatGPT. It was signed by 20,000 people including Elon Musk and Steve Wozniak, but was swiftly mired in controversy and criticism. Still, it arguably prompted a wave of  similar open letters and public appeals all focused on the dangers of AI technology in its current leapfrog form.

In discussion with the newsletter authors last week the founder and president of FLI, MIT professor Max Tegmark, said the point of the letter was really to open the topic for discussion. That reads to me as reframing, but Tegmark does have the relationships in the field and he claims AI researchers and tech CEOs expressed much concern about the “existential risk AI poses, but [feared] being ridiculed as Luddite scaremongers”. He says the letter brought that to the fore, and you can’t fault him there.

But our many anxieties about generative AI don’t seem to have had as much effect on the speed of its development as it has on the language AI firms are now using, heavy on the “trust”, peppered with “ethical”, but not compromising on the powerful promise of innovation.

• Thompson Davy, a freelance journalist, is an impactAFRICA fellow and WanaData member.