AI #05 - How AI Will Transform Businesses
In a previous article, we covered how AI will transform the software industry. Previous articles are here, here, and here. Today, we focus on other large industries for the economy and employment...
Welcome to article 5 of our AI series, where we explore the potential impact of AI innovation on businesses across industries. In a previous article, we covered how AI will transform the software industry. Previous articles are here, here, and here. Today, we focus on other large industries for the economy and employment: Retail, Services, Manufacturing, Financial Services, Healthcare, and Education.
Retail Industry
Amazon disrupted the retail industry, forcing everyone to cut their margins, have an online presence, and offer shipping services. This resulted in many brick&mortar store brands closing, including the iconic J.C. Penney. Shopping malls, except class A and B+, have been struggling post-Covid.
AI offers large conglomerates a new way to streamline their operations. Big retailers like WallMart, Gap, and Amazon employ hundred of thousands of middle class earners in their warehouses. Many of those jobs are set to disappear through robotic. It will go in waves. We are probably looking at 30%-40% in the first wave and over 80% of all warehouse employees after one or two innovation rounds in AI robotic.
In this fascinating podcast, Alex Feldman interviews Robert Playter, Boston Dynamics CEO, about the future of robotics. Robert explains that his firm can create robots that can approximate the mobility, dexterity, and agility of people. Their robots automate routine human activities. Robot lines like Spot and Stretch work for inspection, asset management, construction, manufacturing, public safety, power & utilities, warehouse automation, academia, and research.
They plan to introduce a low-cost new generation of robots. WallMart and Amazon who employ over million people in their warehouses already plan to adopt robots and drastically reduce their workforce. The warehouse industry will see a seismic transformation. An entire pan of middle class jobs will disappear in the US and across the world.
Service Industry
This is another broad industry with many sub-industries. With agents and chat bots becoming the new norm, employment prospects do not look good. Can you imagine robots taking the jobs of waiters, hostesses, and helpers? Think it is far away? Well, think again.
I was traveling through Charles de Gaulle airport in Paris recently. The waiter took my order, and the robot, a tray on wheels, brought my drink. Since Covid, most restaurants have offered their menus with barcodes. My local Chili’s Grill & Bar asks you to place your order through an app on their tablet.
How long before you interact with an AI agent to discuss the menu and order? AI may replace service jobs at McDonald’s, restaurants, and bars. It is not limited to waiters. Tomorrow, robots may replace cooks in kitchens. Most likely, AI will replace low-skilled jobs before moving up the chain.
Manufacturing Industry
This industry is broad. I will just point out a few examples I stumbled upon. Similar to the Retail industry, robots will displace jobs. Once a low-cost robot version is out there, there is no reason to employ an individual who costs more and is not as “reliable”. This is the sad reality. The impact will be less felt in developed countries that have already given up on manufacturing.
There are also positive aspects. For instance, my brother, who leads a maintenance team in a food plant, is looking forward to the day when AI will tell his team in advance what machine or system is at risk of failure. Preventive maintenance can save manufacturing companies millions.
Similarly, AI can do a wonderful job at quality control. The same LLMs used for face recognition or CT scans can be fine-tuned to detect defects in products in nearly real-time. It can greatly enhance the company’s bottom line and protect the brand from bad PR.
I am also excited about the possibilities to optimize the supply chain (inventory, delivery), the production processes (resource usage, new efficiencies, waste management), and workplace safety (workplace accident predictions, improved safety).
Financial Services Industry
The Financial Services Industry was among the early adopters of AI in use cases like fraud detection. If the AI detects a suspicious payment, it can decline the payment in real-time.
Companies like Mastercard, Visa, and Stripe have invested millions in AI anti-fraud detection systems. The incentive is clear: consumers and merchants can use the “Charge back protection” service and get reimbursed for fraudulent transactions. By minimizing fraudulent transactions, credit card companies improve their margins.
AI will disrupt how banks and insurers manage risk. Tomorrow, AI will process your loan application without human intervention. It is a no-brainer. An agent or broker does not add much value. The AI is much better at predicting your risk level and capacity to reimburse the loan based on multiple factors besides your income and assets (age, employment, where you live, martial status, etc.). Over the last few years, banks and insurance agencies have been closing at an alarming rate. AI should accelerate this trend.
In AI 2041: Ten Visions For The Future, Kai-Fu Lee tells the story of The Golden Elephant. It takes place in India. The parents of a family of four bought into a low-cost insurance policy built on an AI-based platform with dynamic pricing based on their profile. The insurance company also offers a suite of apps for investment, home goods, and deals in the area. To get those benefits and special deals, you consent to sharing all your personal information with the company and its partners.
The daughter has a crush on a boy at school. The AI sends subtle messages and makes recommendations to convince her to stay away from that boy. You learn later that the boy comes from a lower caste. The AI is programmed to ignore caste differences, but in this case, it learns from indexed data and implicitly reaches logical conclusions without connecting it to non-ethical and non-equalitarian behavior. Statistically, the AI knows the boy and daughter are not a good match for the model function it tries to optimize.
This specific example made me think a lot. It touches on sensitive topics such as trading your personal data for deals or cheap access to a service (Google, Facebook, Groupons). It highlights the risk for consumers of megacorporations building oligopolies. Finally, it touches on possible flaws in the AI model optimization function. A LLM is a mathematical model optimized for a specific output. There is a real danger of AIs promoting bad behaviors without our knowledge.
Healthcare Industry
The potential for innovation is huge, with expected positive impacts on our lives. Every healthcare professional spends many years learning and becoming an expert in his field. Guess what? LLM healthcare-focused models get ten times that knowledge without decay and can retrain as needed. Will healthcare adopt specialist assistants to help diagnose and provide better care? I sure hope they do, as long as the AI has my best interest in mind and not the healthcare system's (same issue as for insurances above).
A promising field for advancement is radiology. With proper training, a LLM will predict cancer better than doctors. So you may wonder: What use do we have for so many doctors? Good question.
The Montgomery Summit organizes a technology conference every year. In 2023, the focus was on the impact of AI across industries. The AI Revolution In Biopharma gave us a taste of what’s to come. Here’s a summary of the key discussion points:
We are at a “hinge moment” in drug discovery and development process that will transform the industry over the next ten years
Immunology disease researchers leverage big data and machine learning to build large collection of drug molecules (two to ten millions) and test them in a month at a fraction of the cost
In biology, researchers leverage AI to analyze large public datasets, e.g. 180 million immunoacids, to learn about protein sequence and function at a speed and scale that were never possible before
Big Hat uses AI to design better molecules; AI and ML computation let them experiment at a much larger scale; they can now create thousands of molecules every week instead of the average four to eight weeks.
Drug development is transforming from a “bespoke process” to an “engineered process”; it is becoming a data problem where you seek patterns in large datasets
Late stage trials account for 70% to 80 % of drug development costs; 90% of trials do not end in time. AI can drastically reduce that time with digital twins for randomization and blinding
AI models such as AlphaFold and ESMFold perform predictions in protein structure prediction; these models are trained on bigger protein datasets and can predict how specific proteins interact with each other
Having more treatments for the sick is good. Having fewer people get sick should be the goal. Unfortunately, nobody in the healthcare industry is truly concerned about real risks for our health and prevention.
That’s where Medicine 3.0 comes into play. It should not be long until we get our 24/7 medical companion. It will monitor our health, detect anomalies, and recommend the best course of action. This innovation has the potential to save or extend the lives of millions of people.
Education Industry
I shared my initial thoughts here. The biggest opportunity to do good is to offer a teaching assistant to every kid from K5 through K12. Will it happen? Probably not. Or families will be offered the state-approved teaching assistant that will underperform all others. It is probable that new startups will rise and offer tremendous AI apps to assist kids along their learning journey. The assistant will offer personalization settings so parents can define educational guidelines and promote their own cultural and value systems.
In AI 2041: Ten Visions For The Future, Kai-Fu Lee tells the fascinating story of the Twin Sparrows. The twin boys lose their parents and are separated at an early age through adoption. One twin grows up in a wealthy family. His foster father is a successful businessman. He treats the twin like another project and gives him the best AI assistant throughout his education. The goals are extremely ambitious, with no room for distraction. Only results matter. On the other hand, the second twin grows up in a family where he has the space to develop his personality.
Kai-Fu Lee paints a picture of a future where parents with the financial means can buy the best AI and personalize it to control the outcome. Scary stuff!
Some predict that AI will displace white-collar jobs. If this happens, it could create an entire generation of well-educated, jobless, indebted young citizens with gloomy prospects. That one scares the hell out of me. I do not want my kids to be trapped in this system. As a parent, it is never too early to think about your kid’s future career and employability. The sooner you have those hard discussions, the better.
Conclusion
Do we see the same dilemma? AI is going to make our lives better across retail, healthcare, education, and other services. At the same time, it will impact jobs, threatening both well paid middle-class jobs and white-collar jobs. Social levers through education used to be our saving grace. It is unclear whether this will remain the case. Are we seeing two sides of the same coin?
In my next AI article series, I will discuss how AI may transform the workplace.