News not Noise # 20

We are living in the AI century. But it is not evenly distributed. 5 Things we need to do.

Jun 7

Artificial intelligence (AI) is likely the most important technology of our time. It is a general purpose technology that is changing the way we live, the way we work. AI will be key to fight climate change and achieve the UN sustainability goals.

The AI century

The Century of Artificial Intelligence began more than 60 years ago. So we are already in the second half.

In the first half of the AI century, there were many groundbreaking research results. Machine Learning was first discussed in the 1950s. The term Artificial Intelligence was first used in 1956.

The adoption in the real world was mainly driven by the "Good old fashioned AI" in the first half . Based on deduction, AI systems were able to "think" and make decisions. The term used for this: Expert Systems. But it never delivered what it promised. It was very difficult to bring expert knowledge into these systems - because they could not learn from experience.

All that changed in the second half. Driven by the exponential growth of digital data (Internet) and computing power (cloud computing, GPUs) and new insights from research, the old paradigms of machine learning - especially deep learning - are now working in practice. We now have AI systems that can "learn" based on induction - they can learn from data. This has opened up possibilities for real world solutions in many areas - and the use of AI is now growing exponentially.

To be clear, I believe the strength of AI lies in combining it with our human capabilities. Human plus machine will bring us the greatest benefit. AI can bring us "superpowers" if used properly. And we will need these superpowers to combat the climate crisis and develop a sustainable future for us all. This is even more true in uncertain times like today.

For many people and organizations it does not feel like we are living in the AI century.
AI adoption is here but is not evenly distributed

A few facts

Industry adoption

Large organizations are adopting AI at an ever increasing speed. Some of the most valuable companies of the world like Alphabet (Google), Apple, Amazon, Tesla or Facebook are all investing heavily in this technology (same is true for their Chinese counterparts).
Based on the McKinsey report "The State of AI 2021" and the AI Index the AI adoption is continuing its steady rise. 56 percent of the respondents have adopted AI in their organization. And 27 percent saw an increase of their EBIT of at least 5% attributable to AI. (1).

But looking at the adoption of AI in European businesses in 2021 we can see that the average adoption rate is only 8%. Denmark is leading with 24% adoption rate - but most countries are around 10%. (2)

AI Adoption in European businesses. Source: Eurostat

Mainly the smaller organizations are not adopting AI yet. Lack of knowledge, limited access to experts, no visibility of relevant AI solutions might be some reasons we have to address.

Access to computing power

The development of AI solutions needs more and more computing power. Access to super computing is a critical success factor for developing better AI algorithms
The latest statistic of super computing systems and performance shows a very uneven distribution. Few countries have the majority of systems. Developing countries are far behind. (3)

Through cloud computing many organizations and people have access to computing power like never before. But training large AI algorithms requires increasing computing power - and is limited to just a few organizations and countries.

Private Investments

If we look at the private investments in AI - a main driver to develop AI based solutions in startups - we can see a similar picture. A few countries are leading the pace - mainly the US and China (4)

These are just a few examples showing that development and adoption of AI are not evenly distributed - on many dimensions.

5 things we can do

What should we do to change this? Here are 5 things I find particularly important.

Knowledge

Many people still don't know much about AI. We have a lack of understanding that limits the acceptance of the technology. It is important that everyone has a basic understanding of AI, what it can do and how it can be used. This is the foundation for the use of AI in the fight against climate change and for the necessary discussions in organizations, societies and governments about the use and regulation of AI. We should start teaching the basics of AI, algorithms and statistics in schools - and make offerings for all people to get a basic understanding of this important technology. And that's true not just for the developed world, but even more so for the developing world. The Elements of AI course from Finland was a great start (since 2018 translated in 26 languages with more than 780.000 students) (5).
A side effect of greater use of AI is also worth noting: we don't think about AI once we've used it. We talk to Alexa and Siri - but that's not AI to us anymore. We use Google translate or deepl to translate texts - but that's not AI anymore. So in many areas we have already adapted the technology, we don't think about it anymore. AI is always what is not (yet) available.

Experts

Experts are needed to develop AI solutions, and there aren't enough to meet the demand. One reason we are seeing greater use of AI in larger companies: They can afford to pay AI experts high salaries. They can hire the best experts from universities. We need more experts, and many countries have already started to increase investment in AI education. But it's not just about universities. We need to provide access to world-class courses and training to as many people as possible in the developing world. Online courses are one way. Access to the latest research via Arxiv or initiatives like Papers with Code is another good example. Many companies can train their experts internally by investing in on-the-job AI training.

Solutions

In addition to experts, access to AI solutions is a decisive factor for the introduction of AI. Not all solutions need to be developed by experts in the company. Startups and service providers develop solutions for specific challenges. This allows companies to adapt AI solutions without the need for in-house development. However, many companies - especially small and medium-sized enterprises - do not even know that these solutions exist. They face the challenge of finding the best solution available to them - or just understanding that it exists at all. In doing so, they need help. One example: GPAI is developing an SME portal that member states can use to help SMEs find AI solutions for their needs (6). Another example is the "What can AI do for me" service developed in Germany (7).
Investment in AI startups is another critical success factor, as they develop many of the AI solutions. These should be given more support . Especially focusing on startups and solutions with positive impact on sustainability. Finally, we could look to promote AI solutions that can help people in their personal lives - from health to finance, from nutrition to energy - in developed and developing countries.

Resources (Compute, Data, Algorithms)

Access to data, computing power and algorithms are important for the development of AI solutions. Some countries and organizations have a clear advantage here. Open source and open data are important approaches to address these inequalities and enable open access to these resources. Governments, with their purchasing power, can play a critical role in promoting open source and open data solutions. But the same is true for consumers. International collaboration is another important factor. Initiatives such as GPAI or AI for the planet are good examples.

Trust

Finally, trust. Without trust in AI solutions, we will not be able to use this technology to its fullest extent. People need to trust AI-based solutions. Broad knowledge is a basic requirement, but not sufficient. We need regulation of AI, we need more transparency about what AI algorithms are doing. The EU is leading by example with its focus on trustworthy AI (8). But we need to be careful not to over regulate. And we need to get as many people as possible involved in these discussions, as they will be crucial to our future.

Conclusion

To tackle the issues that really matter to us as human beings, such as combating climate change and achieving the UN Sustainable Development Goals, we should ensure that we use our best tools.
AI is on top of this list. That's why we need to address these challenges.

Everyone of us can do something:

Learn more about AI or help others to understand the potential of AI.
Get involved in AI project - even if you are not AI experts. Diversity is key for good AI solutions.
Ask your employer about AI training - or provide this to your employees.
Invest in AI Startups - or use their products.
Support Non-for-Profit organizations to better use AI for good.
Get involved in public discussion on AI rules and regulations.

This list goes on.

(1) McKinsey - The state of AI 2021 -
https://www.mckinsey.com/business-functions/mckinsey-analytics/our-insights/global-survey-the-state-of-ai-in-2021?

(2) Eurostat - AI Adoption 2021
https://ec.europa.eu/eurostat/databrowser/view/ISOC_EB_AI/default/table?lang=en&category=isoc.isoc_e.isoc_eb

(3) TOP 500 - June 2022
https://top500.org/statistics/list/

(4) Artificial Intelligence Index Report 2022
https://aiindex.stanford.edu/wp-content/uploads/2022/03/2022-AI-Index-Report_Master.pdf

(5) Elements of AI
https://www.elementsofai.com/

(6) GPAI - SME Portal
https://gpai.ai/projects/innovation-and-commercialization/broad-adoption-of-ai-by-smes/

(7) What can AI do for me?
https://whatcanaidoforme.com/en/

(8) A European approach to artificial intelligence
https://digital-strategy.ec.europa.eu/en/policies/european-approach-artificial-intelligence