Over the past couple of years, executives have increasingly looked to artificial intelligence and machine learning as part of their digital initiatives. These are increasingly mature technologies with widespread applications for businesses. The new understanding of what AI may bring, is reflected by the results obtained in Gartner’s 2019 CIO Survey, where more than 3,000 executives in 89 countries reported that AI implementations increased over 270% in the past 4 years and over 37% in the past year alone.
“4 years ago, AI implementation was rare, only 10% of survey respondents reported that their enterprises had deployed AI or would do so shortly. For 2019, that number has leapt to 37% – a 270% increase in 4 years,” said Chris Howard, distinguished research Vice President at Gartner. “If you are a CIO and your organization doesn’t use AI, chances are high that your competitors do and this should be a concern.”
Given its huge potential, executives are understandably looking at industrial AI as a game changer. It’s important to mention that AI is a broad concept – it includes tools and technologies, from deep learning, machine learning, and natural language processing, while also including aspects such as robotics and expert systems.
“There are three basic approaches to AI: Case-based, rule-based, and connectionist reasoning”
Marvin Minsky, Co-founder of the MIT’s AI laboratory
Back in 1996-1997, Russian chess grandmaster, Garry Kasparov played a 6 pair match against IBM’s top AI supercomputer, Deep Blue. The IT company spent over $100 million in the project which had 32 processors that could calculate over 200 million chess positions per second in the 3 minutes allowed for a single move in a chess game.
Deep Blue vs Kasparov. The latter not very satisfied with his performance
Source: The Conversation
While this was an important milestone in the evolution of AI, looking back at the “rule-based” approach of Deep Blue (where it searched millions of potential moves), it was relatively unsophisticated in its approach, particularly compared to today’s approaches which often rely on neural networks. However, it serves as a useful example to understand how AI technology has evolved from its nascent stages to become more mature and sophisticated. It is also useful to understand the future of artificial intelligence. Let’s see why:
Computing power has risen exponentially (pretty much according to Moore’s law). Back in the late 90s, an average home computer could run a single core CPU at 100-200 MHz; nowadays, home CPU’s can handle 6 cores at 3.4 GHz. Apart from the speed growth, processing power has become much more accessible.
Evolution of computing power in Floating Point Operations Per Second (FLOPS)
Source: Visual Capitalist
The increase in processing power has been exponential. It’s mesmerizing to think that Apollo’s Guidance Computer from 1969, had the same computing power as 2 Nintendos from 1983.
Increasingly, enterprises are looking to the cloud when building and deploying AI-based applications. Leading cloud computing providers such as Amazon Web Services (AWS), IBM Cloud, Microsoft Azure, and Google Cloud, provide services to enable IT teams to take advantage of the latest developments in machine learning and AI. They also make it possible for organizations to use such services, without having to hire highly qualified data scientists or machine learning engineers.
In the future, instead of using current cloud computing options, we may well see the emergence of decentralized options, like Golem: a blockchain powered open-source decentralized network where people can share their computer power.
Compared to the technology and tools available today, IBM’s Deep Blue was relatively primitive. The AI chess project was built specifically for the task, and used a rules-based approach. The challenge with such early approaches to AI are that it is difficult to scale, and will struggle with more complex situations.
Today, research into AI, and in particular developments in neural networks, have provided a new foundation for organizations wanting to build more sophisticated applications. Practical applications of such “intelligent” machine learning use cases include chatbots and personal assistants. They have already found popular uses, such as customer service agents that generate automated responses on business’ homepages. Chatbots rely on machine learning because it allows them to build conversational data from previous chat “experiences”.
The top benefits of chatbots are 24-hour service (64%), instant responses to inquiries (55%) and answers to simple questions (55%).
Input data is core to the effectiveness of machine learning – for both training data and test data. Machine learning engineers need to have large datasets available with which to train their models. Fortunately, today, businesses have more data available to them than ever before, and as mentioned earlier, can use cloud computing services to store this data at relatively low cost, while also benefiting from significantly greater processing power.
However, it’s not just about the amount of data but the quality of it. As Michele Goetz, principal analyst at Forrester Research states: “Most organizations simply don’t recognize this as a problem.” “When asked about challenges expected with AI, having well-curated collections of data for training AI was at the bottom of the list.”
60% of decision-makers at companies adopting AI cite data quality as either challenging or very challenging—it’s their top challenge when trying to deliver AI capabilities.
Having faster and easier access to computing power, more sophisticated AI tools, and high data availability are 3 key reasons why businesses are finding it easier to adopt AI.
However, the success of your AI initiative will depend significantly on whether you’re able to find highly skilled individuals with expertise developing such solutions. In research conducted by Belatrix, we found that one of the key reasons organizations were struggling to implement machine learning, was due to the lack of skills available.
October 16 / 2019
October 14 / 2019
October 10 / 2019