Artificial Intelligence

The Obligatory Artificial Intelligence Year End Article, 2020 Edition


As we wrap up the year, I’ll start by pointing out something obvious to anyone who reads my column: I’ve rarely mentioned the Covid-19 pandemic. While it has had a major impact on many areas of society, I don’t feel that it has changed the artificial intelligence (AI) and machine learning (ML) markets in a significant way. At most, it has accelerated the adoption of the technology, but it hasn’t done much. While I hope everyone remains as safe as possible, it is not something in my coverage area. My best wishes to all readers remain.

At the end of 2019, my similar column mentioned retail uses of vision and chatbots being more integrated into other applications during 2020. That happened and chatbots have now become a “must have” in most customer interaction interfaces. That has meant a more minimal coverage this year. Retail use of AI vision continues to expand, but in the opposite way of the acceleration mentioned in the first paragraph. The pandemic caused slowdown in retail means spending has slowed while survival is at stake. I expect to see that pick up again in the second half of 2021, as the vaccines become more widespread and consumer confidence improves.

What was exciting about 2020 was the clear evidence of AI tools moving out of the standard worlds of core enterprise data analysis, marketing, retail vision and facial recognition. Two areas discussed in multiple columns were in infrastructure.

The first of those areas is in the cable industry. The large ISPs are focused on better analyzing and optimizing their networks. Smaller companies have come along and are helping the large firms by beginning to use AI at the edge, enhancing modems and routers to better analyze where problems are originating in households. That will improve the experience both for households and the internet providers.

The second was a bundle of segments around facilities. AI is being used to enhance everything from planning through construction and maintenance. One of the more intriguing aspects was the use vision and ML to help businesses and government analyze physical structures in the real world. In the construction arena, AI vision and analysis are helping to better manage and schedule projects.

Discussions with a few vendors shows that artificial intelligence is also working its way into different areas of the sales process. This is a slower change, only adding a bit more accuracy to sales systems in enterprise sales. That will continue, but the real advances I see are coming from the more commoditized markets. Earlier this month, I described the complexity of the commodity channels and sales policies. In this area, I see more visible and rapid adoption of AI to better optimize the channels, while enterprise sales will see a more steady, gradual adoption that is still important.

Government and AI

When looking to 2020 and the future, I still see governments moving slowly. That’s not a surprise. The USA, plenty of other nations, and the EU, have been putting out policy statements, but that’s really all there is. A few statements by people in Congress have shown an increasing interest in the subject, and the Bipartisan Policy Center is looking at it, but still having an Inside the Beltway view of the issue.

Departments are beginning to look at the issue. As healthcare has been an early adopter of AI, especially in radiology, the FDA has begun to look at the issue. A number companies mentioned that the FDA is looking at how to adapt policies to manage changing algorithms, as usual the organization itself isn’t forthcoming. Talking to anyone in the FDA involved in defining the process was not possible, all that was sent to me was vague statements. Seeing how that evolves in the coming year will be interesting.

On the other hand, NGOs are also busy pushing out opinions, policy statements and books. In 2020, I’ve discussed information from the Brookings Institution and the World Economic Forum (WEF). Brookings has some interesting things going on and I’ll continue to watch the evolution of their AI understanding and views. The WEF, being a business group, has almost completely avoided mentioning governments at all. They’ve put out general statements about business ethics with AI, but the regulatory environment will be changing. I don’t expect to see anything significant next year, but I would hope there’s momentum building for action in 2022.

The Tools Themselves

I have constantly returned to the refrain that AI is not a panacea, it’s a tool. One thing preventing wider adoption of the tool is that development software is moving more slowly than I hoped. There have been some user interface changes, with the basics of a graphical user interface (GUI) being layered on some aspects of the development cycle, but there’s still a long way to go.

There needs to be a change similar to that between Third Generation programming languages to Fourth Generation languages, where coding was minimized, and graphical tools made it easier for less technical personnel to build applications supporting their business needs.

One trend I see helping that is that the importance of data privacy has filtered down to the developers. Data cleansing has usually been, sadly, an afterthought. As it becomes more important, and ML uses larger bodies of data, the need to quickly manage that data will drive UI changes. That will hopefully bubble up from managing privacy, to analyzing features for selection, to higher level management of the development process.

Frameworks, such as TensorFlow, as still the main tool for developing ML engines. That requires both more time and more money, as they are complex and require more knowledge and a higher price tag for programmers. While schools are working to meet the demand, there’s also a need to increase the supply by providing more abstract tools that allow more people to leverage the power and promise of AI/ML.



READ SOURCE

This website uses cookies. By continuing to use this site, you accept our use of cookies.