6 hours ago
4
Follow ZDNET: Add us as a preferred source on Google.
ZDNET's key takeaways
- Managing complexity is still beyond the reach of AI.
- For the foreseeable future, humans remain essential for infrastructure and architecture.
- 'Automate last' needs to be the guiding principle in process design.
A funny thing happened on the way to the tech job apocalypse.
The march to AI-driven technology development is hitting a wall -- a wall of complexity. As AI increasingly becomes part of business, it is driving demand for well-designed infrastructure, resilient networks, and sophisticated software stacks that all demand human oversight and intervention.
That's the word from Jon McNeill, CEO of DVx Ventures, former president of Tesla, and former chief operating officer of Lyft. McNeil is the author of a new book The Algorithm: The Hypergrowth Formula That Transformed Tesla, Lululemon, General Motors, and SpaceX. I recently had the opportunity to sit down with McNeill to discuss what IT professionals should consider and look for as they move into this new landscape.
Also: Worried about AI job security? 5 simple ways to pivot at work now (instead of stressing)
For starters, he said, "I'm a techno-optimist, not a pessimist. I'm getting fed up with all the doom and gloom where people are just being half-informed."
Infrastructure and networking opportunities
For technology professionals, robust opportunities are arising out of the AI frenzy, he said. These opportunities differ for infrastructure and networking professionals versus computer science and software professionals.
For infrastructure and networking professionals, demand will be intense, he predicted. "The need for compute, for servers, is creating a ton of demand for networking expertise," he observed. "The expertise needed to keep these servers running, to keep them synched, is extraordinary."
Also: Will AI steal your job? It's complicated, new survey reveals
A significant percentage of GPUs fail each year, so "we're constantly replacing those things," McNeill explained. "When you replace them, you have to re-synch them, and get the networking software working again with the high-band memory chips. All this stuff adds up to big demand for people, and I don't see that going away anytime soon -- with all the complexity of these clusters and server farms."
Along with that, "demand for inference is going to continue to drive demand for infrastructure," he added. "That's really good news for IT infrastructure professionals."
Computer science and software are different story
On the computer science and software side, a different story unfolds, with a call for code writers, software engineers, and developers to move to a higher level of skills, McNeill said.
"They have this layer cake of these different architectural approaches, and smart computer scientists are figuring that out," he noted. "Yes, I can vibe code an app on one single layer. If I have six or seven different models coming together and working together, I can keep them synced agentically. But the invention of the architecture is human -- probably well into the foreseeable future."
The most durable software companies now emerging "are being created with a bunch of layers," he illustrated. "They attack a very complex problem. They'll say, 'Hey, one layer of this problem can be solved by simple search index. Another layer of this problem can be solved by ML.' So I wouldn't spend money on tokens to solve those pieces of the problem. Other pieces of the problem can be solved by small models, maybe another part of the problem by large models."
Also: This OS quietly powers all AI - and most future IT jobs, too
As a result, people "are moving up the architectural value chain higher," he added. "They're letting the basic coding being done by agents and models in QA and QC and deployment. But the more significant architectural work moves up that value chain."
'Automate last' as guiding principle
In his book, McNeill urged proceeding with AI -- and any form of automation -- with deliberate caution and foresight. He related how Tesla was encountering sluggishness in its early factory automation efforts, and was unable to meet demand for its cars. This led his team to conclude that "automate last" needs to be a guiding principle in process design. The company decided to start from scratch in its thinking, setting up a long tent on its grounds with an assembly line operated entirely by humans.
"Our machines couldn't build the cars we so badly needed to sell.... Only after we'd learned the ins and outs of the manufacturing system would we know enough to begin to optimize the process..... The principle of automate last is counterintuitive.... If software is built before the entire system is simplified and optimized, then code is very difficult to change. This is avoidable. Hold the coders off until the end, until you've designed a new, simplified process, optimized it, and know exactly what you want. The coding will go much faster if you have the discipline to wait."
Also: A chief AI officer is no longer enough - why your business needs a 'magician' too
McNeill urged technology professionals to push back if management is demanding AI or high-level, expensive solutions, when a simple approach will work. "Oftentimes, that rationale then convinces people in senior management."
English (US) ·