Software is the key to making machines smarter. It’s at the heart of Facebook’s algorithms and powers smart devices like Amazon Alexa or Google Home and is the key to self-driving cars, advancements in medicine. It’s also critical to banking applications and is the very definition of cyberwarfare. While much of this innovation is good, it can also be used against us if the machines take over, or produce undesirable results.
In the future, we think there are four main overarching trends that point to the greater impact software will have on artificial intelligence and machine learning.
1. Leveraging the compute power of the public cloud
The public cloud has changed the pricing dynamic around two of the most important components of AI and ML which are storage and compute. We believe that public cloud vendors, including AWS, Azure and GCP will get the most leverage from AI and ML given their massive amounts of raw compute power, large data sets and the ability to hire some of the smartest data scientists on the planet.
Each of these three main players have made a substantial investment in the space and have the resources to continue to do so. Although market for AI talent is hotly contested, Amazon, Google and Microsoft have the resources to hire the talent they need for their own AI development stacks on their public clouds. This will bring technology to the masses.
2. Making the most of your data
While public cloud providers look like the winners in AI and ML, most organizations are likely to operate some form of hybrid cloud. So we believe that software solutions that give AI and ML capabilities across both the public and private cloud are likely to attract above-average spending. This is because it gives companies the opportunity to mine for gold as they gain insights from historical data sets that humans could never find. Companies are also using machine learning to automate data discovery to identify areas of risk such as the location of credit card information and the space surveillance network (SSN).
3. What about the application layer?
This is another logical area to leverage AI and ML capabilities as embedded analytics gains additional traction across the application stack. Automation through AI and ML creates efficiencies and offloads repetitive tasks to machines. But we don’t feel AI is a job taker. The combination of AI and humans will increase the ease and effectiveness of cognitive tasks for workers to free up time for them to focus on more mission critical tasks. This makes the workforce more productive and leads to better task prioritization.
- AI augmentation is forecast to recover 6.2 billion hours of workforce productivity while generating $2.9 billion in business value by 2021 according to Gartner.
4. Security and AI/ML – a perfect match
The 2015, the Global Information Security Workforce Study by Frost & Sullivan predicated a shortage of 1.5 million workers in the cybersecurity sector by 2020. While the Centre for Cyber Security Safety and Education’s 2017 study expects this gap to be 1.8 million by 2022.
This has become a real issue for the industry, so companies are investing in security software solutions to address the labor gap. By automating repetitive tasks, cyber-security professionals are free to focus on network security and to respond to events. The most common uses of AI and ML are to prevent targeted and zero-day attacks by using existing data to identify the unknown, eliminate alert fatigue by decreasing the number of false positive security events and ranking and prioritizing potential events. And will AI take away jobs for IT and cyber-security personnel? We don’t think so.
- Artificial Intelligence could generate 2.3 million jobs by 2020 according to Gartner.
Four sectors where AI is transforming industry and society
The four major disruptive trends in the automotive industry enabled by AI are connectivity, autonomous driving, shared mobility and electrification. The more miles the AI takes in, the more it learns, so miles = knowledge. Just as the smartphone created a whole new eco-system of business, so too can the autonomous vehicle. Autonomous driving improves safety and mobility and is an economic opportunity to promote a culture of the shared. And Robo-taxis could be a reality by 2025.
Tangential industries could be affected by this trend as insurance premiums fall but the companies are also likely to experience a lower number of claims. Autonomous vehicles could provide more opportunities for entertainment and advertising as well as the telecom industry. The drinks industry might see benefits as impaired driving becomes less of an issue. On the flip side, what happens to car dealerships, auto retail and repair and public transportation?
- Technology could bring down the cost to the consumer of sharing an autonomous vehicle to nearly 50 cents a mile.
The future is likely to be less bio and more tech as drug developers team up with AI companies. AI and machine learning can transform how drugs are developed from early scientific discovery through clinical trials to commercialization. This process currently costs around $2.6 billion, with approximately $1.1 billion of this sum spent on preclinical testing. As only 14% of potential new drugs in the clinic make it to approval, AI-driven approaches could help identify the relevant drugs needed to target disease more effectively. Precision medicine could also cut medical costs by using AI to improve diagnosis and prevention and reduce medication failure.
Applying AI to therapeutic modalities such as gene editing can also augment the drug development process by enhancing key drug characteristics as tech companies become increasingly important to biotech-AI endeavors.
These improvements to traditional drug development strategies and applications should lower the failure rate as an increased R&D efficiency expands the operating margins for biotech companies.
Indeed, the rise of precision medicine, leveraged by AI, could ultimately decrease the societal economic burden of medical treatment by diagnosing patients earlier, reducing the number of failed medications a patient cycles through, choosing safer more effective medications for a patient’s specific disease, and disease prevention through improved diagnostics. We anticipate a number of start-ups and partnership to emerge alongside up-and-coming therapeutic technologies to better harness these powers.
At the forefront
- Gilead Sciences (GILD) recently partnered with Alphabet’s Verily Life Science to leverage machine learning to better understand immunological diseases and molecular signatures.
As an industry, banking is traditionally slow to innovate, but an article in Innovation Enterprise published in 2017 suggested that 75% of current banking operations could be carried out by robotic process automation. Already AI is streamlining middle and back-office functions as digital takes over from paper while banks believe the biggest potential for AI-related savings are in the front office. Many banks are already using bots or virtual assistants such as Citigroup’s Facebook messenger chatbot. However, the biggest challenges for banks will be around the disconnect of data from legacy systems and the mountain of data still held on paper.
AI is also helping banks to fight fraud. Denmark’s Danske Bank was struggling with a fraud detection system which threw up 1,200 false positives each day, 99.5% of which were not fraud related. They used AI to realize a 60% reduction in false positives (expected to increase to 80%) and a 50% increase in true positives, freeing up resources to focus on actual fraud.
And while trading will become more automated, with the use of facial-recognition programs, this area is a challenge for AI because the rules are constantly changing. Nevertheless, AI-based trading companies that are deep technology-focused will continue to grow although the market is likely to consolidate with fewer players gaining more market share. The AI-powered Exchange Traded Funds space is also expected to grow.
- A Business Insider article in 2017 predicted that robot-advisers for products and services will manage around $1 trillion by 2020, $4 trillion by 2022.
Autonomous applications will play a larger role in mining operations as AI technology develops. Already in Australia Rio Tinto are using GPS-controlled self-driving vehicles the size of a two-story building to improve safety and efficiencies. While in Sweden’s Lapland, Volvo are testing an autonomous truck, which can’t be guided by GPS, in its underground Kristineberg Mine.
By 2025 more sophisticated hardware will mean nearly all homes in the United States will have smart meters with advanced metering infrastructure (AMI). This will give utility companies real-time updates on damage, or potential damage (before a storm), to any equipment while customers will have greater control over how much energy they use.
Taking AI to the next stage
The marriage between the physical and the digital is being called the 4thIndustrial revolution as AI capabilities power machine learning, smart factories, advanced automation and intelligent supply chains. This is the next stage after the industrial Internet of Things and data analytics. We’re still many years away from transforming and optimizing manufacturing processes and creating radically new business models as Industry 4.0 becomes mainstream.
At the forefront
- GE ‘s ‘brilliant factory’ initiative will create facilities that house an ecosystem of smart robots advanced sensors and computers that communicate with each other in real time to ensure pristine quality and the smooth-running of the production line.
The impact of AI on the workforce
Because AI has wider implications for every area of human experience, not just the economy, we’ve limited our analysis in this report to Narrow AI because of its huge potential to destroy jobs. Indeed, it’s likely that disruptions to the labor market will be similar to those seen historically in farming and manufacturing. The AI-age has the potential to upend the current order and lead to a de-globalization of sorts. And there will be winners and losers as AI, not the labor force, becomes the dominant driver of economic growth.
Farming peaked as a percentage of the labor force in 1820, but it took nearly 100 years for the level of employment to peak and roll over. This suggests a symbiotic relationship between human labor and machines in the early stages of adoption. But given the faster pace of technological innovation, the time between the share of employment and employment level peaks will be much shorter.
For areas of the world with a slowing population growth, the proliferation of Narrow AI in the labor market will be a blessing. While European and Japanese economic growth has slowed over the last decade, these areas are well-positioned for the AI-age. But Africa and Asia (mostly India) will add more than 400 million people to their working population over the next decade and nearly double that over the next 20 years. How to employ all these people will be a challenge.
- The United States auto sector has the highest use of robotics – more than 1,200 robots per 10,000 employees compared to less than 100 for all other manufacturing – yet the effects of automation have been quite modest.