For 2023, AI Growth Still Going Strong
2023 trends: personalization, large language models, AI and cars, generative AI, layoffs make top talent available, hybrid NLP, multimodal learning, human insight AI
By John P. Desmond, Editor, AI in Busines
The growth of AI continues to be strong, with worldwide spending on AI software expected to surpass $500 billion in 2023, according to IDC market analysts, a growth of 20 percent in the past year.
“Over the last decade, Artificial intelligence (AI) has become embedded in every aspect of our society and lives,” stated technology writer and analyst Bernard Marr, in a recent account on LinkedIn. “It’s hard to ignore its impact,” he stated.
The community of readers of this AI in Business newsletter published via Substack submitted their own predictions for where AI will have an impact in 2023. A selection of their thoughts follows.
Roberto Masiero, SVP of Innovation for ADP, the human service software provider:
“In 2023, we will see personalization becoming more and more ubiquitous across all disciplines of the world of work. We'll use data, machine learning (ML) and artificial intelligence (AI) to create models tailored to individuals. With that, we'll be moving from an environment in which HR is a ‘world of rules’ to HR becoming a ‘world of exceptions.’ We'll see a diversity of behaviors when it comes to the relationship between the employer and the employee.”
Varun Ganapathi, CTO and cofounder of AKASA, developing AI solutions for healthcare operations:
Large language models (LLMs) will help people be more creative: We’re likely going to see more solutions like GitHub Co-Pilot where large language models (LLMs) are able to assist people in meaningful ways. These tools are not going to get everything right, but they will help solve that initial writer's block. When you’re staring at a blank page, it's often hard to get started. But if you can describe the prompt or problem and the model outputs something, it may give you a good starting point and show you something you can use — even if it’s not entirely what you want. Prompt engineering (i.e. instructing models using the right starting text) will become a new way of writing programs, using natural language.
.It may be that AI will actually help us become more creative — by seeding us with initial ideas that we can build upon and refine.
Natural language processing (NLP) + object recognition will bring search to the next level: While most people write scrapers today to get data off of websites, this may soon be replaced by further advancements in NLP. You will just have to describe in natural language what you want to extract from a given web page, and it will pull it for you.
For example, you could say, "search this travel site for all the flights from San Francisco to Boston and put all of them in a spreadsheet, along with price, airline, time, and day of travel." It's a hard problem, but we could actually solve it in the next year.
As another example on the healthcare side, I think we'll be able to predict — automatically — the notes and documentation a doctor might write for a given diagnosis or treatment, which would be a huge achievement. It could save healthcare workers valuable time.
Generally, we’ll be able to tie object detection — where we train algorithms to predict what is in an image — to natural language processing. This would be a big step forward as it will allow us to simply describe what output we want, and it would figure out how to build a classifier to deliver it. For example, you could say “does this image contain an animal with four feet and tails?” and that would be “programming” a classifier.
While we can do this to some extent now, it will become more advanced in the coming year and allow us to go one level deeper — only describing attributes of what we want to find rather than providing labeled examples of the object itself. We may also develop new methods for combining prompt engineering and supervised labeled examples into a coherent whole.
Businesses leveraging AI to do more with less during challenging times will win in the long term: Microsoft CEO Satya Nadella recently said, “software is ultimately the biggest deflationary force.” And I would add that out of all software, AI is the most deflationary force.
Deflation basically means getting the same amount of output with less money — and the way to accomplish that is to a large degree through automation and AI. AI allows you to take something that costs a lot of human time and resources and turn it into computer time, which is dramatically cheaper — directly impacting productivity.
While many companies are facing budget crunches amid a tough market, it will be important to continue at least some AI and automation efforts in order to get back on track and realize cost savings and productivity enhancements in the future.
Thomas Ma, founder and CEO, Real Messenger, a social app for real estate agents, operating in Los Angeles and Hong Kong:
The death of iBuying in real estate. The current ibuyer models are taking advantage of sellers/buyers, stealing agent’s leads and then making profits out of it. High transparency builds confidence for buyers/sellers looking for the best agent. Honesty and trust are the keys to the future of proptech.
AI and data analytics aren’t only for property assessments. They have multiple uses. Agents can use insights to evaluate and learn their content performance, client behaviors and to improve their business model and marketing strategies.
Potential buyers and sellers are open to trying new technology to bring them more accurate and customized listings. The MLS is too slow and listings sell before they hit the market. Buyers and sellers are looking to online tools to get ahead of the game. Even if it means giving away their data for free.
Sujatha Sagiruaju, CPO, Appen, supporting machine learning data development:
AI Will Create The Car of The Future: The automotive industry is expected to grow from $5.6 billion in 2018 to $60 billion in 2030. Moving forward, AI will play a significant role in transforming vehicles, particularly when it comes to safety and consumer experience.
.In the next year, AI will make large strides in safe driving technology for autonomous vehicles (AVs). The European Union launched the General Safety Regulation in June 2022, which mandates safety technologies such as distracted driver protection, lane-keeping systems, advanced emergency braking and pedestrian collision warning, to be included as standard in new vehicle types. The EU expects to save 25,000 lives and avoid 140,000 serious injuries by 2038.
Innovation in AVs will not be exclusive to safety, though. As AI spending in the automotive industry increases, AI has the opportunity to improve the consumer experience inside the vehicle. Manufacturers are already using AI to create features such as in-car voice assistance and Tesla’s Autopilot, which allows you to summon your vehicle from its parking space. In the next year, we will see improvements within the in-cabin experience geared toward comfort, such as auto-adjusting seats, automatic sun glare protection and more personalized infotainment systems.
Nicholas Harris, CEO and founder of Lightmatter, the photonic supercomputer company:
As Generative AI and its resulting content explodes, the energy consumption and load on hardware chips will only increase and accelerate. Generative AI is taking the internet by storm. Now publicly accessible, new AI technology, such as AI-generated portraits and artwork, is proliferating across social platforms. This sudden increase in access and popularity will have a direct impact on the energy consumption load placed onto hardware chips. Already, current AI models running on traditional chips require massive computing power from data centers. Given the demand for AI and ML, its compounding annual growth rates, and the ever-increasing need for compute power, we’re already facing a crisis. In 2023, hyperscalers will need to factor in this energy consumption as AI and ML continue to see explosive growth through content creation of generative AI.
Top tech talent will start and invest in the next generation of AI start-ups: The second half of 2022 brought a wave of sizable layoffs across the technology sector. As this cohort of tech talent searches for their next opportunity, I expect to see a wave of new startups founded and invested in the new year. As AI becomes more accessible to the masses and its use cases permeate every industry and sector, the next year, and even next five years, will create the next generation of disruptive startups. As in any time of economic uncertainty and change, the companies that break through are the ones that will create lasting futures and impact.
Jeff Catlin, CEO, Lexalytics, an InMoment Company focused on text analysis and natural language processing software:
AI goes ROI: The slowdown in tech spending will show up in AI and machine learning in two ways: major new AI methodologies and breakthroughs will slow down, while innovation in AI moves toward "productization.” We'll see AI get faster and cheaper as the innovation moves into techniques to make deep learning less expensive to apply and faster through models like DistilBERT, where accuracy goes down a bit, but the need for GPU's is reduced.
Growing acceptance of hybrid NLP: It's fairly common knowledge that hybrid NLP solutions that mix machine learning and classic NLP techniques like white lists, queries and sentiment dictionaries mixed with deep learning models typically provide better business solutions than straight Machine Learning solutions. The benefit of these hybrid solutions means that they will become a checkbox item in corporate evaluations of NLP vendors.
Paul Barba, Chief Scientist, Lexalytics:
The rise of multi-modal learning: The wave of image-generating networks like Stable Diffusion and DALL-E demonstrate the power of AI approaches that understand multiple forms of data - in this case, image in order to generate a picture, and text in order to take in descriptions from a human. While multimodal learning has always been a significant research area, it's been hard to translate into the business world, where each data source is difficult to interact with in its own way. Still, as businesses continue to grow more sophisticated in their use of data, multimodal learning jumps out as an extremely powerful opportunity in 2023. Systems that can marry the broad knowledge conveyed in text, image and video with sophisticated modeling of financial and other numeric series will be the next stage in many companies’ data science initiatives.
The singularity in our sights?: A research paper by Jiaxin Huang et al. was published this past October with the attention-grabbing title "Large Language Models Can Self-Improve." While not yet the singularity, the researchers coaxed a large language model into generating questions from text snippets, answering the self-posed question through "chain of thought prompting," and then learning from those answers in order to improve the abilities of the network on a variety of tasks. These bootstrapping approaches have historically had a pretty tight bound to improvement - eventually models start teaching themselves the wrong thing and go off the rails - but the promise of improved performance without laborious annotation efforts is a siren song to AI practitioners. We predict that while approaches like this won't drive us into a singularity moment, it will be the hot research topic of 2023 and by the end of the year will be a standard technique in all state-of-the-art, natural language processing results.
Dr. Rana el Kaliouby, Deputy CEO, Smart Eye, supplier of vision systems, and former CEO and co-founder, Affectiva, offering emotion recognition systems:
Synthetic data will accelerate AI innovation: In 2023, synthetic data will be a game-changer in accelerating the development and deployment of AI, while guarding against algorithmic bias. One of the significant challenges in developing AI is getting the right amount and diversity of data to train machine learning-based algorithms. These algorithms require massive amounts of data that are representative of the different people that will interact with it and the contexts in which it will be used. It is difficult, time-consuming and costly to acquire this breadth and depth of data. Data synthesis enables AI companies to rapidly augment their existing datasets and simulate scenarios that are difficult to generate in the real world. For example, in automotive, synthetic data tools can use a source image of a driver to create synthetic variations that use varying lighting conditions or head movements. It could even simulate a driver falling asleep behind the wheel – data that is rare and very dangerous to capture in real life. Deploying synthetic data tools is key to not only solve these complex challenges of data collection, but also to combat algorithmic bias, by ensuring datasets are truly diverse.
From unimodal to multimodal AI: Already, Human Insight AI – technology that understands, supports, and predicts human behavior in complex environments – has been deployed across a number of industries, from automotive to healthcare. However, up until this point, most of these deployments have used a limited number of sensors, each collecting data in a silo. In 2023, we are going to see the development of multimodal Human Insight AI, where a multitude of sensors can come together to give more accurate predictions and deeper insight into the complexities of human behavior. In healthcare, multimodal digital biomarkers can accelerate the development of precision and preventative medicine to make more accurate diagnoses and provide early interventions. We’ll also see multimodal Human Insight AI continue to enhance industries like automotive. For example, cars will be able to use multiple, inward-facing sensors to capture data from eye tracking, facial analysis, and vital signs like heart rate, heart rate variability and respiratory rate. Fusing data from these different sources will provide a more holistic view of what’s happening with occupants in a vehicle and create a safer and more comfortable in-vehicle experience.
(Write to the editor here.)