Some of the Latest Trends in Artificial Intelligence
We’re into our second year of publishing a “Global AI Race” series of articles on artificial intelligence startups from around the world and it continues to pose a challenge. We use an objective measure of “total funding taken in so far” and that excludes any firms that choose not to disclose funding or are bootstrapped. We search for various categorizations like “artificial intelligence” or “deep learning” and that means we’ll miss any firms that haven’t chosen those categories in their Crunchbase profile. But the ones we worry about the most are those firms that we might include in one of our “top AI startups” lists that don’t actually do AI. It’s a huge problem, and one that was highlighted recently by a European venture capital firm, MMC Ventures, that surveyed 2,830 startups in Europe that were classified as being AI companies and found out that 44% of these companies were incorrectly classified as being “AI startups.”
Still, the fact that there are 1,580 AI startups across Europe means that we’re reaching a tipping point, or what MMC Ventures calls “a divergence.” Working together with Barclays, they produced a 149-page report titled “The State of AI: Divergence 2019” which takes a holistic look at AI across the globe finding a growing division between leaders and laggards. We pored through every page of that report to extract some of the latest trends in artificial intelligence that you might find insightful.
Artificial Intelligence and Adoption
“In 2019, AI ‘crosses the chasm’ from early adopters to the early majority,” says the report. As an investor, you should now be looking for companies that don’t use artificial intelligence since they’re quickly going to become laggards. For many of the companies adopting artificial intelligence, it may seem all new and shiny. Truth be told, AI has been around for decades with “seven false dawns” taking place between 1965 until now.
Artificial intelligence may be “the fastest paradigm shift in technology history.” In just three years, the number of enterprises with “AI initiatives” rose from 1 in 25 to 1 in 3. One in ten enterprises use more than ten AI applications, and the most popular use cases are chatbots (26% of enterprises), process automation solutions (26%), and fraud analytics (21%). Nearly half of all companies prefer to buy AI solutions from third parties as opposed to building their own.
Globally, China leads the charge with twice as many Asian firms adopting AI as compared to North American firms. The report points out some interesting high-level reasons why China has become a global leader:
- Data Availability – China has more permissive policies than Europe regarding use of personal data.
- Less siloed data within companies – According to MIT Sloan Management Review, 78% of leading Chinese companies maintain their corporate data in a centralized data lake, compared with 37% of European and 43% of US companies.
- Legacy technology – Chinese companies typically have fewer legacy applications and processes to deal with.
It’s not surprising that two out of three reasons involve data. The best AI algorithms are the ones with exclusive access to high-quality data sets. With that said, some of the developments being made in artificial intelligence hardware and technologies are of equal importance.
Artificial Intelligence Technologies
When it comes to understanding the underlying technology behind artificial intelligence, most of us can get by with the very basics. For the people who are building these algorithms, it’s a different story. They’re highly paid and highly educated. Salaries for AI engineers average $224,000 at the 20 highest-paying companies and 60% of AI developers have a Master’s or Doctoral degree. Demand has been so high for talent, that even the academics are being pulled into the corporate world. According to The Economist, between 2006 and 2014, the proportion of AI research publications including an author with corporate affiliation increased from approximately 2% to nearly 40%.
There are more than 15 approaches to machine learning, and trying to understand even one would probably take as much time as studying for the CFA with about the same benefits at the end. Instead, you just need to know about the latest trends in machine learning so you can throw them around at your next board meeting to demonstrate what a thought leader you are. Here are some trends to watch when it comes to how artificial technology is developing.
AI Hardware Trends
- From GPUs to Custom Silicon – Firms are now building custom silicon chips with high degrees of specialization in favor of GPUs like the ones built by Nvidia and AMD. We covered some of the startups building these chips in our piece on 12 AI Hardware Startups Building New AI Chips.
- Edge Computing – Edge computing is where we move away from reliance on the cloud and empower devices ‘at the edge.” Custom silicon helps move AI to the edge “by coupling high performance with low power consumption and small size.”
AI Software Trends
- Generative Adversarial Networks (GANs) – Have you seen the realistic photos of people created by AI? They were actually created using GANs. As the name implies, the method operates with two networks – a ‘generator’ and ‘discriminator’ – working in opposition to create increasingly lifelike media. GANs can be used for a wide variety of other applications such as data normalization, network security, system training. For an excellent detailed description of GANs, see page 80/81 of the report.
- Reinforcement Learning (RL) – The AI algorithm is presented with a goal and experiments through trial and error while being rewarded for progress towards the goal. Requires no human intervention. Great for applications that lack training data. It’s how Google’s DeepMind mastered Go.
- Transfer Learning (TL) – Emerging method that uses skills learned from a previous problem, and applies them to a different but related challenge. Interest in TL has grown seven-fold in the last two years. 2018 was a breakthrough year for the use of TL in Natural Language Processing (NLP).
- Artificial General Intelligence (AGI) – It’s not in the report but something we thought should be in your vocabulary. It refers to the point where artificial intelligence becomes as smart as a human with intelligence that’s broad and adaptable. It’s where what we’re trying to accomplish as we adopt techniques like RL and TL.
It’s to be expected that a technology being adopted so quickly will also be developing as fast. The report also makes mention of how quantum computing will impact AI, something that we talked about in our article on Artificial Intelligence (AI) and Quantum Computing. Let’s look at how AI is being applied in the real world.
Artificial Intelligence Applications
Teachers and managers will be somewhat reassured to find out that their jobs will be “more resilient to AI in the medium term” while others won’t fare as well. The report talks about how AI will have the greatest impact on job functions where a majority of time is spent collecting and making sense of data.
Based on the above, it’s not a surprise to learn that the sector with the highest adoption of AI firms is the insurance sector. Nearly half of all insurance companies have deployed AI or plan to in the next 12 months. (Check out our piece on How Technology Will Affect Big Insurance Companies.)
Insurance is a good example of an industry where AI brings both threats and opportunities. One threat is the autonomous vehicle which impacts the 42% of global premiums that come from car insurance. According to Autonomous Research, UK car insurance premiums are expected to fall by as much as 63%, causing profits for insurers to fall by 81%. These losses might be partially offset by improved fraud detection. According to the Association of British Insurers, fraud costs UK insurers around $1.7 billion a year, something they currently spend around $265 million annually trying to prevent.
It’s easy to see how fraud detection can impact the bottom line, but what about chatbots? They’re the most pervasive “application of AI” with more than 20% of enterprises now using them.
Just because you plug in a chatbot to your corporate site, does this now mean you can say your firm is “using AI?” The question to ask is how much the chatbots are impacting the bottom line. What’s incredible to see in the above chart is how drastically these answers have changed in just one year. Of the 2,600 firms surveyed in 2018, 72% didn’t use any AI applications. Ask roughly the same number of companies the same question one year later, and the situation has changed dramatically. The number of uses cases for AI has expanded dramatically, and the report cites 31 core use cases across eight sectors.
Seems like it was just yesterday when we were looking at the promise of AI applied to medical imaging. Today, 40% of healthcare providers use AI-powered computer-assisted diagnostics.
AI is about data, and loads of data will be generated by the Internet of Things (IoT). Out of all sectors, the utility sector has climbed on board with IoT the most. According to Gartner, 67% of all utility companies now use IoT technologies such as sensors – all of which are generating loads of big data that can then be fed to AI algorithms. Across all sectors, we’ll see a move towards “X-as-a-Service” business models that will service the preference of enterprises to buy vs. build. These business models also serve to centralize the data so that the AI algorithms will become better and faster through economies of scale.
European Artificial Intelligence Startups
Since the report is written by a European VC, it’s no surprise to see that they’ve carved out the European AI startup space by country and presented it in this handy chart.
We’ve covered a number of these AI startups in our “Global AI Race” series, and were surprised to see that about a third of Europe’s AI startups can be found in the UK. Zee Germans and the French have developed their own AI ecosystems and the Spaniards should be noted for punching above their weight. Nine in ten of Europe’s AI startups are business-to-business (B2B) which means they sell solutions to enterprises as opposed to consumers. That’s changing now, as a quarter of new AI startups are focused on selling solutions to consumers.
Six in ten European AI startups are early stage (Angel or Seed-stage funding.) Here’s a good quote from the report on how early-stage AI companies can make for compelling acquisitions as larger firms look to acquire or risk being passed by:
AI startups are valuable suppliers – an ‘on-ramp’ to AI – for companies that embrace them, while disrupting those that do not.
Alongside these success stories will be high-profile failures as well. Then we get back to thinking about all the startups that are incorrectly classified, whether intentional or not. Today, one in twelve European startups put artificial intelligence at the core of their value proposition. There’s good reason for that too. Forbes published an article on this topic titled Nearly Half Of All ‘AI Startups’ Are Cashing In On Hype which explains why:
Startups labelled as being in AI attract 15% to 50% more funding than other technology firms.
That’s why it’s so important for investors to do their own due diligence prior to investing in companies that claim to be using AI. That’s exactly what the firm that produced this insightful report excels at.
MMC Ventures states that AI is a core area of research, conviction, and investment for their firm. In the last several years they’ve made 20 investments, comprising 50% of the capital they have invested, into many of the UK’s most promising AI companies. We’re going to look at some of those companies in coming articles and deep dive into what’s happening in the UK AI startup world.
As investors, we expect the impact of artificial intelligence on the bottom line should be reflected in earnings over the coming years. In some cases like the insurance industry, this impact could be negative. When the dust settles, artificial intelligence will become something that companies can no longer use to obtain a competitive advantage because everyone else is doing the same thing. At what point does AI move from disruptive technology to commodity?