Main content

PTC LiveWorx - the first rule of AI Club is talk about AI!

Chris Middleton Profile picture for user cmiddleton May 18, 2023
Summary:
AI was conspicuously low key in yesterday's keynote, but PTC has its own story to tell.

An image of someone pointing with their finger to a screen with bright blue digital circles

After an impressive, sensory-overload of a keynote – which came across as much as a mothership landing as a corporate presentation – PTC President and CEO Jim Heppelmann had a more difficult close encounter. A small room heaving with analysts and journalists, rather than a rapt hall full of customers and partners.

While his keynote’s real message had been “We’re back! Isn’t it great to mingle with customers and partners again?” (it’s life, Jim, and just as we remember it), Day1’s management team Q&A was more a case of rebuffing analysts’ obsessive questions.

Flanked by former ServiceMax CEO Neil Barua, now President of PTC’s Service Lifecycle Management (SLM) division, and CTO Steve Dertain, Heppelmann was probably hoping for a media high-five at what, in reality, has been a very well-staged, successful, popular event for the $15.35 billion product-management company. 

But frustratingly for the trio, the assembled inquisitors really wanted to know why there had been so little focus on AI when it has been dominant in everyone’s corporate messaging this year. As previously reported, Heppelmann’s response to diginomica’s own question about the wisdom of focusing on the industrial Metaverse at the expense of PTC’s AI strategy, was:

A lot of stuff ended up on the cutting room floor [from the keynote script]. So, I was probably just sparing you from [a presentation that would have taken] even more time. We've talked a fair amount about generative in the past, so maybe others just felt like it was a new idea. But again, we had to cut the session back, so it was a cutting on the floor.

However, a simple statement about how generative elements in PTC’s CAD tools can already aid users would have sufficed at the keynote, plus a positioning statement about the evolving potential of AI in analysing data from the integrated portfolio of SaaS offerings that now includes ServiceMax. 

So, it was perhaps inevitable that questions about GPT, large-language models, and generative AI piled up at the Q&A, despite the lack of relevance in some analysts’ lines of inquiry. Evidence that buzzwords really are everything in the IT industry, much as both diginomica and PCT might wish otherwise. 

Remember - tech companies don’t just make applications, hardware, and services, they also sell memes more successfully than anyone else. It’s why ‘the cloud’ somehow describes massive data centers built on land, despite being the least appropriate term imaginable. So, at industry jamborees, the absence of a popular buzzword becomes more significant than it should, and thus starts a narrative of its own that CEOs are forced to engage with. 

There is a story

For PTC’s leadership team, the panel at least presented a good opportunity to stress what the company can already do with generative tools. And on that point, Heppelmann was keen to set the record straight:

“I want to come back to generative design, because CK [Catherine Kniker, PTC’s Chief Sustainability Officer] made that comment in the keynote about us generating 3D content. And she said that is conceptually like ChatGPT [the only significant mention of AI in the 80-minute keynote presentation].

But the thing about ChatGPT is it’s so easy: you put in text, and it gives you text back. But, it's hard to express 3D in text. In fact, the only way to really express 3D is in 3D. So, the way generative works in [PTC’s design tool] Creo is you give it a 3D concept and it fleshes out the concept and gives you back a pretty good answer. So, it's really generated 3D – 3D in, 3D out. It’s a little different to GPT, but it's still AI-based generative technology.”

PTC has been offering this facility for some time. Heppelmann added:

ChatGPT largely took off because it's so democratic. Any of you can just Google ‘ChatGPT’ and pretty soon you’re clicking on a link and trying it.

Heppelmann suggested that PTC is exploring ways in which generative 3D models could be made available to the public experimentally, so that the same viral adoption curve we have seen with GPT, Mid-journey, Stable Diffusion, and others, could take place. In the meantime, the likes of Adobe are already infusing AI across all of their imaging tools, of course.

But there would be problems in deploying AI in other sections of the business, he added – especially when it comes to reusing code. Heppelman says: 

The problem is, AI may learn from copyrighted code, and from malware. So, the great news is AI is very productive at generating code. But the bad news is some of it might be copyrighted, and some of it might be malware, so we have to be careful. But we are investigating it.

An untested area, of course, is whether any company could copyright code generated by an AI system. That aside, CTO Dertain fielded yet more questions about GPT, saying:

Every product that we have is a great candidate. So, for example, in ServiceMax there's an AI chatbot built within it [ZINC] to help field-service technicians [more on this in our exclusive interview with Neil Barua]. Certainly, putting a little bit more contextual information in, we can get GPT to understand how that data is related, which gives users a much higher-quality experience in terms of how we deliver service. Then lather, rinse, and repeat the same pattern for [product lifecycle management tool] Windchill, and for all of our data-centric products that create a large corpus for language models. 

But GPT is not so much related to Creo directly. I don't know how I'm going to have a conversation about the geometric design side of things, for example. But there are other forms of AI that we're looking at inside of design itself, not just what we're already doing with generative [with generative design elements in PTC’s CAD tools].

Expect us to bring in more of that in future. There is an active project – six or seven of them, in fact, that we are juggling at the moment. Also, in conjunction with Microsoft, we are really working on the accuracy of things. It’s often too loose, and we don’t want it to serve up the wrong answers to customers.

[Other uses might include] making documentation easier to read, especially across multiple languages. Again, teams are looking at using it, at how our products present themselves, and at automation’s benefits to the team.”

Then he added, perhaps more out of frustration than anything else:

For all of us in the audience, we should probably take generative AI, the large language model, the OpenAI stuff, and just separate it [from generative design elements in CAD]. They are very different, extraordinarily different. But, I do believe OpenAI-style capabilities can lead a person through a conversational UI experience via prompt engineering. And so, when we set up a generative design, I can see those two worlds colliding.

Heppelmann added:

My own view of this stuff is that of a co-pilot concept – helping my engineer. I think we're quite some distance from saying, ‘We don't need the engineer anymore, because the AI does everything for us’. My view is, today, all about making engineers more productive. That’s where we’ll get to a good answer quick.

My take
As noted, I suspect that the absence of an AI focus in the theme-setting keynote had a longer tail than the management team intended, forcing the issue to be discussed later. 

But an alternative perspective came from an anonymous guest at the event, a representative from a healthtech company. That person, speaking privately to diginomica over lunch, said of generative AI’s relevance to most industries:

The real customer use cases just aren’t there yet for the enterprise. What we really wanted to know after four years of COVID hiatus, is how do we do the digital transformation stuff that we should have been doing all this time?

So, it seems Heppelmann had the right idea all along.

 

Loading
A grey colored placeholder image