The Future Today Institute (FTI) has published its 15th Tech Trend Report, in which it provides a comprehensive overview of technological trends that will change our everyday lives and work. The report comprises over 650 pages and provides a well-structured overview of technological trends in a wide range of areas – digitalization, artificial intelligence (AI), 3D printing, drones, blockchain, and more.
The trend report is based on the evaluation of extensive data on consumer behavior, but also on insights into the activities and successes of numerous research institutions. Based on this quantitative basis of the trend report, a reliable short- to medium-term perspective emerges. In some places, the book breaks out of the tight medium-term time frame and looks further into the future in optimistic as well as pessimistic scenarios. Incidentally, this is reminiscent of the development of various scenarios around artificial intelligence (AI), which the founder of the FTI (Amy Webb) has played through in her book The Big Nine
In the following paragraphs, I present some selected facts from the Tech Trend Report 2022. I highly recommend reading the free report; the report is structured in small reading bites, so that you can browse through it from time to time. For each chapter (13 in total) there is an overview of particularly relevant trends and innovations.
Below is a selection of the most exciting trends and developments around AI (=Chapter 1); for each of the other 12 chapters I’ll provide some “information bites” as an “appetizer”.
Click here for the Download: Tech Trend Report 2022
Artificial Intelligence
AI has long since arrived in the creative scene. From music composition to painting. An interesting insight into the dynamics of this development is provided by the AI-based program GauGAN, which was developed by the chip manufacturer Nvidia for image composition . The following video (2 min) shows how the program translates user specifications into photorealistic landscape shots:
Even more interesting or challenging in the field of image generation is the DALL-E project from OpenAI. DALL-E was developed last year. This is an AI model trained to manipulate visual concepts through speech. The starting point is a prompt or image description in natural language, based on which a series of images are generated that reflect an interpretation of the linguistic description.
Below is the generated image material for the following text prompt: “an armchair in the shape of an avocado, an armchair imitating an avocado”. The images are taken from the following website: DALL· E: CreatingImages from Text.
It’s worth mentioning the advances in language models, which have made headlines over the past year. There was another breakthrough here in 2021. To measure language comprehension, there is the so-called General Language Understanding Evaluation (GLUE) benchmark: The baseline value for humans is 87, and between May 2018 and August 2020, AI systems for natural language processing improved from 60 to 90.6, already surpassing humans. The SuperGLUE benchmark is a new measurement with more difficult speech comprehension tasks. When SuperGLUE was introduced, the gap between the best model and humans was almost 20 points between the best-performing AI model and the human performance on the leaderboard. Last year, Microsoft and Google’s AI models outperformed human performance, as predicted by the FTI, by the way.
This improvement in the performance of language models is achieved (amongst other factors) by ever larger language models. GPT-3 set a benchmark at the beginning of 2021 with 175 billion parameters. But this one is already outdated: Google’s Switch Transformer and GLaM models have the dizzying number of 1 trillion and 1.2 trillion parameters; but even bigger is the Wu Dao 2.0 model from the Beijing Academy of AI, which reportedly has 1.75 trillion parameters .
And it’s not just AI models that continue to create superlatives; this also applies to hardware: Today’s neural networks required enormous computing power for a long time, took a lot of time to train, and relied on data centers and computers that consumed hundreds of kilowatts of electricity. That’s all starting to change. We are talking about the SoC, the System on a Chip. In 2019, Cerebras unveiled an AI chip with 1.2 trillion transistors, 400,000 processor cores, 18 gigabytes of SRAM, and interconnects (tiny connection nodes) capable of transmitting 100 quadrillion bits per second. That’s a staggering amount of components and performance — and yet the company announced last year its next-generation chip, the Wafer Scale Engine 2 (WSE 2), with over 2.6 trillion transistors, 850,000 cores, 40 gigabytes of on-chip memory and memory, and 20 petabytes of memory bandwidth.
Where there is so much efficiency, the question naturally arises in which areas this is used in economic life. The Future Today Institute (FTI) has conducted its own modelling and research on this topic. The FTI sees a leap in productivity through AI in the following areas in particular:
sources: “Occupational Employment and Wage Statistics,” U.S. Bureau of Labor Statistics; ” The Impact of Artificial Intelligence on Labor Productivity,” Eurasian Business Review; “Economic Impacts of Artificial Intelligence (AI),” European Parliament; “A Future That Works: Automation, Employment, and Productivity,” McKinsey Global Institute; Reuters Institute, Oxford University
AI researchers now also see realistic development paths towards general artificial intelligence: Researchers are developing individual algorithms that can learn multiple tasks. DeepMind, the team behind AlphaGo who learned how to play Go at the level of a human grandmaster, continues to push his research. MuZero mastered several games without being told the rules, a “significant step forward in the development of general-purpose algorithms,” according to DeepMind. In a groundbreaking paper, “Reward Is Enough” (late 2021), DeepMind researchers hypothesized that artificial general intelligence could be achieved through reinforcement learning alone.
Capabilities or areas of application of AI are also increasingly advancing into areas that make humans more readable, predictable and manipulable. For example, through the recognition of emotions: A new type of neural network can determine how people feel. Using radio waves, AI can detect subtle changes in heart rhythm, perform pattern analysis, and analyze a person’s emotional state at a given moment. A team from Queen Mary University of London used a transmitting antenna to bounce radio waves off test subjects and trained a neural network to detect fear, disgust, joy and relaxation while showing them various videos. The system was able to correctly assign emotional states in 71% of cases.
Amazon‘s Rekognition API, on the other hand, detects a person’s emotions based on facial recognition and physical appearance. Replika, on the other hand, uses AI to evaluate speech and text and reflects the user over time. Affectiva Human Perception AI analyzes complex human conditions using speech analytics, computer vision, and deep learning. For example, the automotive sector uses Affectiva‘s technology to detect a driver’s emotional state – such as drowsiness or anger on the road – and provide real-time suggestions to improve driving styles.
scientists also warn that advertisers may end up changing and controlling buying behavior through “sleep and dream hacking.” And: “Sensory clickbait” is on the rise, which aims to manipulate users’ emotions.
There is no doubt that some use cases and areas of application are ethically highly critical; Regulations and ethical guardrails are increasingly being discussed. Several governments, for example, are trying to regulate deepfake technology. Bills have been introduced in California, Texas and Massachusetts to regulate or ban deepfakes, and a number of federal bills are currently being discussed.
In other areas, governments are pushing ahead with the use of AI, which is ethically controversial to say the least: In 2021, the U.S. military stated that it had begun using AI to control its airstrikes by using algorithms in a live kill chain. The kill chain is a process of gathering information, conducting analysis, weighing risks, and using weapons to destroy a target. An AI system was inserted into the Air Force Distributed Common Ground Common Ground System; the AI system can’t order an attack on its own, but it now automatically identifies possible targets.
Incidentally, the following applies: The chapter “Artificial Intelligence” (approx. 60 pages) in the Tech Trend Report does not provide a conclusive overview of innovations based on AI; Rather, as a basic technology, artificial intelligence basically permeates all areas, and in all subsequent chapters you will also find trends and innovations that are AI-driven. However, the first chapter provides a good overview of the state of research and core applications such as language models.
metaverse, home of things, robotics, blockchain, drones, and more
For the next 12 chapters, I’ll limit myself to naming the headline for the chapter and give some exciting development trends as an “appetizer”:
Scoring, Recognition & Privacy: In the 14th Tech Trend Report, the authors declared as bad news: “Anonymity is dead. And: Everyone alive today is being scored.” COVID-19 has further destroyed expectations for privacy. Employers’ monitoring of workers has accelerated as the pandemic has normalized remote work.
metaverse, AR/VR, synthetic media: People create multiple digital versions of themselves, each tailored for specific purposes. This will lead to fragmentation – and an ever-widening gap between what a person is in the physical world and the people they portray on different online platforms. With Diminished Reality (DR) consumers will soon be able to visually and audibly block out what they want – and whom – in real time.
Work, Culture & Play: Gaming has become the engine of an entire cultural and commercial ecosystem.
News & Information: Natural language search interfaces – whether in AI assistants or as a function in browser-based search engines – are threatening the marketing strategy of many media companies.
Health & Medicine: Telemedicine is becoming more and more comprehensive and start-ups are expanding into new areas of medical care. Breakthroughs in sensors and artificial intelligence in particular are expanding the possibilities of remote diagnostics. Technology also makes it possible to carry out laboratory work at home.
Home of Things: HoT platforms are increasingly serving entire households rather than just one person. As a result, consumers will be more willing to embrace electronic devices that work seamlessly in their smart home. Homes are also becoming sentient: automated systems detect and regulate temperature, sound, light and other functions in real time to support families.
Policy, Government & Security: Digital ID systems already exist in some countries. Canada’s SecureKey is a blockchain-powered identity system used to mediate trusted interactions between financial institutions, governments, and telecom network operators. Citizens can connect various online services with a digital credential. For banks, digital ID can reduce fraud, improve the security and privacy of customer data, streamline authentication, and automate transactions. In Estonia, all citizens, regardless of where they live, must have a digital identity issued by the state.
Logistics, Robotics & Transportation: “By 2025, 52% of current workplace tasks will be automated —DHL’s “Future of Work in Logistics” report.
Decentralization & Blockchain: Visa has had the monopoly on payments at the Olympics for 36 years, and in 2022, China’s e-CNY was the preferred means of payment within the Olympic community. Basically, China is pushing for wider use of its digital currency, the e-CNY, inside and outside the country.
Telecommunications & Computing: The first steps towards building a quantum internet are already underway. The global race for quantum computers is in full swing. Several countries, including the U.S., France, the U.K., and China, aim to become the world leader in quantum computing.
Synthetic Biology, Biotechnology & AgTech: New DNA testing kits offer personalized recommendations for recreational drugs such as marijuana and MDMA. And soon, we may be able to write megabytes of data per second to synthetic DNA that will be readable for thousands of years.
Climate, Energy & Space: Several new geoengineering initiatives will be tested in 2022, including cloud injection (salt particles are injected into low-lying clouds to make them more reflective), sinking iron into the oceans, and altering solar radiation.
