BREAKINGVIEWS-Amazon summit melds AI hype to corporate drudgery

By Sebastian Pellejero

LAS VEGAS, Dec 5 (Reuters Breakingviews) - From casino magicians to bar-room Elvis impersonators, Las Vegas has long played host to an uneasy meld of fantasy and dreary capitalist reality. As 60,000 coders, consultants and technology evangelists descended on Sin City for this week’s 14th edition of Amazon.com’s AMZN.O sprawling cloud computing conference, that tension only grew. The festival of techno-optimism on the Strip reflected an industry convinced of an imminent artificial intelligence revolution, yet still struggling to turn that promise into day-to-day corporate practice.

For all the cheer, the snags are plain. Attendees ranging from creative software giant Adobe ADBE.O to cybersecurity provider Zscaler ZS.O trumpet billion-dollar AI ambitions, yet surveys suggest over 80% of experiments in real businesses never graduate to real-world use. Fewer than a fifth of workloads have moved from company servers to third-party data centers, Matt Garman, boss of Amazon’s cloud unit AWS, told the host of the "Acquired" podcast on stage at the Venetian. This is the industry’s core bind: rolling out cutting-edge AI depends on computing infrastructure that many firms still haven’t adopted.

The revelry continued regardless. Between marathon keynotes and packed technical demonstrations were Formula 1 simulators, giant slides and tattoo stalls, while mascots angled for selfies. High-paid engineers queued for free haircuts and Intel bucket hats. The pageantry contrasted with the slower work of getting AI to function inside real companies.

Amazon’s silver bullet is decidedly practical. To tempt corporations onto its cloud servers, it needs to offer a clear payoff. To that end, the conference’s repeated mantra was that firms should adopt agents — systems that combine an AI model, software, and task-specific tools to perform multi-step work with limited human oversight. The aim is to make AI predictable in corporate settings, converting hype into hard returns.

In contrast to the most far-flung promises from AI boosters of curing disease or unleashing wild new frontiers of economic growth, the pitch is surprisingly tangible. Cox Automotive, a car-dealership software provider, said agent-driven tools have cut insurance-estimate times from two days to around 30 minutes. Spaceship builder Blue Origin now relies on agents for, among other things, simulating lunar dust. The PGA golf league boasts of using agents to produce ten times more web content.

These examples suggest that, even as AI labs tout progress cracking complex mathematical proofs or researching new materials, the machine learning market will resemble the conventional business-to-business software market. Venture capitalists already seem to sense it, judging by the explosion of investment in call-center automation companies. Whether that justifies the vast sums of cash being burned by the likes of OpenAI is unclear, but it at least offers a plausible story pointing to rising business adoption. For once, what happened in Vegas is unlikely to stay there.

Follow Sebastian Pellejero on LinkedIn.

CONTEXT NEWS

Amazon.com hosted its 14th annual cloud computing conference, AWS re:Invent, in Las Vegas from December 1 to 5.