Occasionally it is helpful to step back from the day-to-day and review your business, firm, industry or situation with a longer time frame. Sometimes we can only assess where we have been by doing so.
The exercise arguably is more difficult when trying to extrapolate where we are going. “Predictions are hard, especially about the future,” many, including physicist Niels Bohr, have quipped.
“Extrapolations from techno-scientific innovations have a distressing capacity to be deterministic,” says historian Amanda Rees. One example might be the impact of computing and communications evolution on social, economic, political or scientific endeavors.
It is easier to describe and predict some changes in computing capability than to predict how they might affect changes in the biological or social world.
Still, to the extent that any specific problem can be solved if sufficient computing power is available, at low cost, there are at least some indicators of potential.
Many of us might note that we are able to use millimeter wave radio frequency spectrum for consumer and business communications only because the cost of signal processing--enabled by the reduction in computing cost, form factor, along with increases in capability--allow us to do so much sophisticated signal processing that the spectrum can be made to work for consumer communications.
I have argued in the past that an understanding of Moore's Law “saved” the U.S. cable TV industry in the 1980s when high-definition television was developed.
Perhaps we might also say that those same developments in performance made possible streaming video services that now are cannibalizing cable TV.
The point is that it is difficult to extrapolate future developments in a linear way from linear improvements in computing capability. But it sometimes helps to think about the application of computing in situations where business models formerly unthinkable can become quite practical.
Anything we see in consumer internet applications--where capabilities are supplied at no cost to users--provides an excellent illustration. The classic question is what does your business look like if a key cost constraint is removed.
Though we might have mischaracterized key elements of the argument, ride sharing did raise questions about what it would mean if “cars were free.” They obviously are not “free,” buit musing about changes in personal transportation have happened because of the existence of ride sharing.
The difficulty always is that other drivers of behavior also exist. Consider consumer demand for mass transit, which seems to be falling as other options--and social changes--develop. Many riders had less need--or no need--during Covid-19 pandemic restrictions on “going to work or school.” But lower mass transit ridership trends were in place even before Covid, both internationally and in the United States.
But many speculate that the availability of ride sharing has diminished use of public transportation, though other social forces also seem to be operating.
Likewise, we might argue that vastly-improved computing and storage price-performance curves are good enough to allow applied artificial intelligence in a growing range of use cases. Most of those use cases involve inferences about future impact based on historical metrics.
Letting farmers know when to water or apply fertilizer, and in what quantities, should lead to improved crop production. Industrial processes likewise should be improved when we can predict when a particular machine will fail, or what must be adjusted in real time to optimize output.
Lots of other supply chain or process processes likewise should benefit from cheap and ubiquitous ways to manage and optimize present flows of resources, whether that be people walking, cars on highways or other logistics-related issues.
Computing progress means new applications or use cases can develop in a non-linear way, even when computing rates of development are linear.
Technologist Ray Kurzweil noted in 2005 that “in 1968, you could buy one (Intel) transistor for a dollar. You could buy 10 million in 2002.”
Looking at the cost of a single compute cycle, Kurzweil also noted in 2005 that “the cost of a cycle of one transistor has been coming down with a halving rate of 1.1 years.”
“You get a doubling of price-performance of computing every one year,” he said.
One likely impact in the global communications industry is the impact of AI-assisted networks on worker skill requirements, to say nothing of the improvements in network performance or availability.
Some argue that skills will need to be upgraded as networks get smarter. The countervailing argument is that skill requirements might change, but not as much as people think. When the networks are smarter, they will be able to predict potential outages or degradations, allowing automatic or manual changes to prevent problems.
Outside plant or core networking work might not become more complicated at all; the work might become less complicated, in terms of adjustments and maintenance. That might shift possible priorities in other ways that involve different tasks and skills, though not necessarily “higher” skills (depending on how one defines “higher).
Perhaps more effort shifts to marketing and away from plant maintenance. That might involve different skills, but not necessarily “higher” skills for most people. When social media algorithms dictate marketing actions, the heavy lifting is done by the algorithms.
People at firms simply need to know what outcomes they wish to achieve.
No comments:
Post a Comment