by Jean-Louis Gassée
As the sudden explosion of public interest in ChatGPT continues to excite millions, we ask: Is this the tipping point for machine-driven conversation (and more)? Is ChatGPT the Netscape of our time?
In Fortune’s The inside story of ChatGPT: How OpenAI founder Sam Altman built the world’s hottest technology with billions from Microsoft, author Jeremy Kahn helpfully explains OpenAI’s history, structure, financing, and much more — at 6K words, the article covers a lot of territory. Kahn cuts straight to The Big Moment scenario in his opening paragraph [emphasis mine]:
“A few times in a generation, a product comes along that catapults a technology from the fluorescent gloom of engineering department basements, the fetid teenage bedrooms of nerds, and the lonely man caves of hobbyists — into something that your great-aunt Edna knows how to use. There were web browsers as early as 1990. But it wasn’t until Netscape Navigator came along in 1994 that most people discovered the internet. There were MP3 players before the iPod debuted in 2001, but they didn’t spark the digital music revolution. There were smartphones before Apple dropped the iPhone in 2007 too — but before the iPhone, there wasn’t an app for that.”
(Unfortunately, the rest of Kahn’s piece is behind a paywall that offers various Fortune subscription incentives, a fairly common and understandable practice. I’ll just note that a single click on a widely available browser on desktop and mobile devices allowed me to download the article — and read it three times.)
For this week’s Monday Note, I queried a number of serious technologists, AI gurus with recognized expertise in both academe and the Valley about the possibility of ChatGPT as a “Netscape Moment”. What Saith Our Noble and Worthy AI Elders about the surprisingly powerful and friendly artificial conversationalist that was developed in the open (literally)? Will this new widely-available technology change the landscape?
I was expecting skepticism. These individuals aren’t easily inflamed by the latest Metaverse, Augmented Reality, or Autonomous Driving predictions. So I was surprised that that they were more than cautiously optimistic; indeed, they (mostly) anticipate a Netscape Moment — and jobs displaced.
I inquired about ChatGPT’s imperfections, felicitously called “hallucinations”, when the engine comes up with wrong answers. Obvious mistakes are amusing and innocuous, but the hallucinations become dangerous when the error is difficult to eyeball. I got two responses.
First, this is with the current GPT-3.5 version; an upcoming GPT-4 release will take care of the problems. Not a great perspective, one that brings back perennial FINR (Fixed In Next Revision) stories.
A more convincing reply combines the inventive, discursive ChatGPT output with the less voluble but more reliable answers of a search engine. As one (coffee) drinking companion suggested, this could be what Microsoft has in mind for future evolutions of its Bing search engine and other applications, or what Google search might do with Generative Machine Language (GenML) extensions.
Another “fixing ChatGPT” suggestion can be found in Stephen Wolfram’s recent paper titled Wolfram|Alpha as the Way to Bring Computational Knowledge Superpowers to ChatGPT. Over the years (I recall conversations with him about 35 years ago) Wolfram has developed powerful computing technology made available at wolphramalpha.com.
No doubt, ChatGPT’s popularity, its “virality” will give rise to more such additions to its powers, especially if the OpenAI/Microsoft association actively markets ChatGPT licenses.
We shouldn’t neglect what established AI companies have to say.
Google, for example, calmly publishes a blog post titled Google Research, 2022 & beyond: Language, vision and generative models. Detailed and informative, the blog posts remind us of the depth and breadth of Google’s AI work and gingerly refers to ChatGPT without mentioning it by name [as always edits and emphasis mine]:
“2022 has brought exciting advances in media generation. Computers can now interact with natural language and better understand your creative process and what you might want to create. This unlocks exciting new ways for computers to help users create images, video, and audio — in ways that surpass the limits of traditional tools!
[…]
2023 and beyond will surely be marked by advances in the quality and speed of media generation itself. Alongside these advances, we will also see new user experiences, allowing for more creative expression.
It is also worth noting that although these creative tools have tremendous possibilities for helping humans with creative tasks, they introduce a number of concerns — they could potentially generate harmful content of various kinds, or generate fake imagery or audio content that is difficult to distinguish from reality. These are all issues we consider carefully when deciding when and how to deploy these models responsibly.”
This contrasts with wild rumors of turmoil in the Alphabet executive suite with “code red” alerts and founders emerging from their warrens (while laying off about 12,000 people). Let’s hope Impressive Research will soon translate into Product Innovation for the rest of us.
From Alphabet’s reserved stance we now move to Meta’s Chief AI Scientist Yann LeCun, an individual with internationally acclaimed credentials and (you be the judge) an attitude to fit. A ZD Net story titled ChatGPT is ‘not particularly innovative,’ and ‘nothing revolutionary’ recounts LeCun’s reaction [again, edits and emphasis mine]:
“The public perceives OpenAI’s ChatGPT as revolutionary, but the same techniques are being used and the same kind of work is going on at many research labs, says the deep learning pioneer.”
He then adds faint praise…
“It’s nothing revolutionary, although that’s the way it’s perceived in the public, it’s just that, you know, it’s well put together, it’s nicely done.”
…reminds us that ChatGPT is using AI research and technologies developed by others…
“You have to realize, ChatGPT uses Transformer architectures that are pre-trained in this self-supervised manner. Self-supervised-learning is something I’ve been advocating for a long time, even before OpenAI existed…
Transformers is a Google invention, unveiled by Google in 2017, which has become the basis for a vast array of language programs, including GPT-3.
The work on such language programs goes back decades.”
…and explains that everything will be forgotten when Meta offers a surpassing alternative to ChatGPT:
“Are we going to see this from Meta? Yeah, we’re going to see this. And not just text generation, but also creation aids, including “generative art. I think is going to be a big thing.”
Unmentioned in LeCun’s commentary is Blenderbot, an August 2022 Meta chatbot that promptly got roasted for various racist and imbecilic responses.
We might very well be at a Netscape Moment, whether from OpenAI and its partner Microsoft, or perhaps from countermoves by Alphabet or Meta. OpenAI may have gotten there first, but let’s recall that Netscape didn’t come to dominate the browser market it ignited.
One last imprecise, lingering thought after my third reading of Jeremy Kahn’s Fortune article and (unsuccessfully) trying to parse the unusually complicated terms of the Microsoft investment deal. Admiring how deeply smart Sam Altman is, I couldn’t help but wonder how much advantage he might be taking of Satya Nadella’s thirst for a perceived AI coup.
— jlg@gassee.com