Hello, fleshy carbon-based lifeforms — welcome to the AI future!
AI is all the rage in tech, financial markets, and the next wave of the future, lately, but what’s the buzz? We’re here to break down and dissect exactly how it’s disrupting markets, what the business model looks like, and how you might benefit from (or be replaced by) AI.
It’s one of the biggest fears in a lot of people’s minds when it comes to AI: will it make your role obsolete?
A recent report by Goldman Sachs states that, “artificial intelligence (AI) could replace the equivalent of 300 million full-time jobs.” That’s about the entire population of the United States as of time of writing. Given how quickly so many jobs are being replaced by AI, it helps to be flexible about how to both integrate AI, letting it help you, rather than having it replace you.
So buckle up and let’s get started.
7 ways AI is already here
First things first: AI is already here. Here are seven examples in no particular order:
Platforms like Spotify, Netflix, and YouTube use AI algorithms to suggest new music and videos based on your preferences.
Social media platforms
Facebook, Instagram, Twitter, and other platforms employ AI to recommend content for your feed and display targeted ads, for better or worse.
Facial recognition and photo apps
Apps like Adobe Photoshop and Lightroom use AI to recognize faces, scenes, and objects, making it easier to edit and enhance photos. Facial recognition is also used for unlocking devices and accounts.
Navigation and rideshare apps
Google Maps, Waze, and Apple Maps utilize AI for real-time traffic updates, optimized routes, and predictive analysis to provide accurate travel information. Uber and Lyft’s algorithms also assign drivers to customers with AI, matching based on rating, distance, and a variety of other factors.
Virtual assistants and chatbots
Virtual assistants such as Siri, Google Assistant, Alexa, and Cortana assist us in our daily tasks using AI-powered conversational capabilities. We’ve also been running into their precursor: automated customer service on both phone and chatbots using rudimentary AI to deliver quick answers and fixes to customers at odd hours.
Google Translate uses AI to provide real-time language translation, helping users communicate in different languages.
Apps like Mint, Robinhood, and Acorns leverage AI to offer investment suggestions and help users manage their budgets effectively.
It’s highly unlikely we’d go through modern life without touching at least one of these — imagine going back to a world where we navigated driving to a new place with written or printed instructions!
That leads us to our next point: what about AI and the future of work?
AI in the workplace: strengths
If AI is already making our lives easier in small and subtle ways in our personal context, it only makes sense we’d want to incorporate it in work.
Dr. Louis Hyman pointed this out about ChatGPT: “It’s not the end of work, it’s the end of boring work.” He points out that the majority of us don’t work at Google or Apple, but at companies that have yet to go paperless. Some companies might eliminate roles and solve the exact same problems with less people with the onset of AI, but he believes that AI will effectively free people up from tedious or difficult tasks (especially those around data manipulation) to go on and solve more complex, more human tasks.
That brings us to some of the principles around where AI’s best suited:
This is the go to: any tasks that can be automated because they’re repetitive and time-consuming. This could include things like:
Data entry and processing
Using OCR (Optical Character Recognition) tech, you can scan documents in bulk, rather than processing data one sheet at a time.
Factories will have AI test and inspect products for defects en masse, rather than having somebody check one by one.
Everything from social media posts to interviews/office meetings can be scheduled ahead of time and automated with AI, with emails or notifications sent out ahead of time. We probably already do this, without taking into account the complex series of steps that happen afterwards (a recurring meeting will send invites and notifications to the necessary parties).
Large set data analysis
AI’s great at analyzing large amounts of data for a variety of reasons: whether it’s making sure to redact all sensitive information from several terabytes of data before disposing it, or offering any recommendations or insights after parsing through years of reports, AI is built to look through data in the way that no human can do without mistakes or tedium settling in.
Email sorting and filtering
Our inboxes are already doing this to some extent, with spam filtering. With a little work, we can also set up our emails to either highlight them in specific colors or send them to specific folders.
Microsoft, having inked a multi-billion dollar deal with OpenAI, the creators of ChatGPT, is rolling out Microsoft Viva.
If you work in an office, you’re likely using Microsoft Outlook, which comes equipped with Viva Insights at the top of your options under the Home menu.
If you haven’t clicked it yet, you’re in for a treat - it sorts through your email inbox to highlight all the upcoming meetings and priority emails that may be important, as well as suggested tasks that you may have missed and require doing.
Personalization and recommendation
When you sign onto any account — from Spotify to your work email — no one person delivers hand-picked songs to you or specifically calls out the emails most important to you. When it’s a software that has thousands of users, a lot of the personalization options come down to taking in data points about how you use the software and then fine tuning the software to work for you.
From auto-completing sentences in written emails or auto-correcting spelling issues, to cloud storage apps automatically sorting and categorizing files based on naming conventions to settings, to even just having light/dark mode turn on at the right time based on the time, we’ve got plenty of neat automated personalization settings to tinker with.
If you haven’t yet, consider tinkering with those settings to get the best experience possible.
If you’re working in a data-heavy white collar industry that involves crunching numbers, from advertising to healthcare to retail, you’d likely find yourself managing industry specific tools known as business intelligence. Whether it’s finding trends in treatment to analyzing retail sales data to organizing KPIs, business intelligence and data visualization tools like Tableau, Microsoft Power BI, and Looker. All of these have functionality that will point out insights or suggestions based on data ingested.
Physically strenuous or unsafe conditions
The future of robots and AI in our workforce started in conditions that were too dangerous or simply unreachable for humans without specialized equipment. The recent Oceangate scandal where five people died in a submersible highlights one of the most hazardous conditions: deep sea. For years, it’s been imagined that AI-powered robots would operate in conditions of extreme heat, pressure, and danger would be the future. AUVs (autonomous underwater vehicles) could repair underwater cables — yet these aren’t a reality just yet.
Instead, we’re seeing robots closer to home. Amazon is already testing out new fully autonomous mobile warehouse bots as well as limited use of flying drones to deliver packages. The other condition that physical robots powered by AI can solve for is not just physically stressful situations, but also situations filled with tedium and boredom. Work situations that are long, hard, repetitive leads to employees that end up making mistakes in the end.
AI in the workplace: shortcomings
Before we start trying to figure out the best way to insert AI into every single aspect of our work, consider these shortcomings of AI.
Garbage in, garbage out; any AI that requires feeding data into it. Yann LeCun, one of the AI experts at Meta, points out that LLMs (large language models) will still ‘hallucinate’ when it doesn’t have the appropriate data: it’s the evocative term describing the phenomena where, given the lack of concrete data sources, AI can make up things.
RLHF, or reinforcement learning through human feedback, is a lot of where the quality for ChatGPT comes from — thousands of humans grading responses on ChatGPT until it integrates the data and regurgitates something consistently acceptable.
AI algorithms will have a hard time with certain common sense questions (the kind, for example, that will list apples and oranges, but only ask about apples).
Consider the case where a lawyer submitted a legal brief written by ChatGPT. Often in situations where ChatGPT is asked to come up with sources, the generative AI will often come up with citations that sound like a facsimile of something passably close to reality, but be fabricated hallucinations instead.
Rather than saving work, AI can generate more work as sources are required to cross-check these citations instead.
Limited creativity and organizational structure
AI — especially the most recent generative text AI — has a few limitations by design.
First, whatever writing AI generates often has a blandness to it that’s the result of several cycles of human testers reinforcing acceptability and non-toxicity. While great for summarizing concepts in a concise way, AI is less creative when it comes to varying sentence structures and writing something more interesting. Whatever generative AI creates is by requirement a result of whatever data and reinforcement was put into the model.
Second, AI can show mediocre results when it comes to storytelling or stringing along narratives that make sense from beginning to end. It has no way of relationally connecting the dots, so the narratives can come off as fairly simple. It’s far better at rewording content or spell-checking, which necessitates that it takes one set of data (inserted content) and runs it against another (its existing LLMs).
Privacy and data security
Depending on exactly what you’re doing or who your employer is, unleashing AI on sensitive data can be an issue, especially when the AI you’re dealing with is licensed to a competitor. Abuses of large amounts of data have been evident in everything from the Facebook Cambridge Analytica scandal down to Robinhood’s selling of retail trading data.
Before you start integrating AI into your workflow, it’s probably a good idea to check with your company to make sure that sensitive information is guarded carefully, and all security/privacy protocols are followed.
AI in human resources
Considering all the strengths and shortcomings of AI in the workplace, we can point to one field where it’s already being integrated at a high level: around 40% of human resource departments today utilize some form of AI in their workflow.
The process of recruiting is a strenuous process with multiple stakeholders and multiple moving parts. At the end of the day, while people’s livelihoods are incredibly important to them, the recruiting role is one where nobody’s dying (unlike high risk situations like heart surgery or space travel) if there are failures, and any improvements welcomed would have reverberations throughout every organization.
Here are some of the ways AI are being integrated already in HR practices:
ATS — applicant tracking systems
ATS systems are already in use in the vast majority of large corporations, using bots to a) build out job descriptions, b) scan through resumes to find percentage match options for roles and c) track progress overall of job applicants as they move through the system to detect issues like organizational biases and/or hire better candidates over time using machine learning.
On the flip side, a quick search of AI resume apps shows plenty of options on the market that take job descriptions and build out tailored resumes for jobs to meet the ATS system where it is, like Rezi.ai, Resumai.com, KickResume, and Resume.io. In a hyper-competitive hiring environment, getting past the ATS is a huge part of the challenge.
Interview meeting scheduling
High level hires at corporations typically require several rounds of interviews with different executives and/or hiring managers, after the preliminary recruiter screening. In the past, the methods of lining up tight schedules between the stakeholders who need to show up and the recruit, with vastly varied schedules, was a process of Tetris: taking blocks of time that don’t necessarily line up and trying to force an overlapping fit.
These days, simply having each individual’s calendar input into HR scheduling systems can generate a list of suitable times, through apps like Microsoft’s Scheduler. It takes out the guesswork, but still likely requires somebody to double-check time slots.
AI productivity applications like Clockwise can integrate everything from project management applications to scheduling and email applications to arrange schedules and make work organization a little smoother.
Challenges of Implementing AI in the Workplace
So why isn’t everybody implementing AI?
The IDC (International Data Corporation) pointed out in 2019 that, for a quarter of companies, half of all projects to integrate AI into their enterprise solutions fail.
If it was easy, everybody would be doing it. There’s a few reasons why, but it comes down to cost and training people on AI.
In many ways, AI is very expensive. Between the processors required to run AI (especially the content generation AI), to the power required to run them, to the massive number of people required to reinforce learning (see Alexandr Wang, who built a billion-dollar business from providing and training the labor to teach AI systems) to the expensive data required to jump-start an AI (quality information doesn’t come from just anywhere) — the whole process is resource-hungry.
What that means is, sometimes, it just doesn’t make sense to try to use AI or automation to fix something that might be a one-off. The ratio of cost to usefulness quickly diminishes when the applications aren’t general. Because AI has traditionally been built to solve repetitive tasks, it’s the much wider applications of LLMs (large language models) that inspire so much interest in recent editions of ChatGPT.
Upskilling and reskilling an aging workforce
Upskilling and reskilling are some of the key issues with implementing AI; with an existing workforce, without extensive training, it might be tempting to go back to the way it was. Simpler things break less often, and if things work, it might not make sense for companies to build around new technology with an aging workforce.
In creative industries like design, AI is outclassing designers with years of experience in terms of productivity and quality of design; what they don’t have, however, is necessarily the eye to detect the best of a variety of designs.
A potential scenario is the oldest of the workforce unable or unwilling to upskill/reskill will have a hard time, but the seasoned middle-to-senior professionals will transition into managing and curating AI products through experience.
What AI might do is completely eliminate a lot of low level production work that juniors and entry level roles are currently doing. It may also apply pressure so that only exceptionally talented juniors will make it in an already tough industry.
AI applications in different industries
ChatGPT has proven to be extremely strong at generating code for software engineering. ChatGPT can very effectively generate code that is as good, if not better than human software engineers. With this tool, a software engineer who is strong at seeing connections can generate the nitty-gritty code of individual parts, but string them along to solve complex issues.
This takes the busy work out of coding, leaving engineers to strategically solve the big picture problems while QAing whatever output code ChatGPT comes up with.
Applying machine learning algorithms to medical imaging data like X-rays, MRIs, and CT scans has yelped to detect diseases like cancer earlier. AI is also able to comb through massive amounts of patient data and both highlight the most important points, as well as detect trends in larger patient populations, for the sake of predicting outcomes and optimizing treatment plans.
Marketing and communications
AI can quickly generate ad copy personalized to specific demographics, allowing rollout of campaigns requiring specific targeting much more quickly. Business Intelligence applications like Tableau and Microsoft Power BI allow marketers to generate insights on the fly and share them quickly.
Much of digital marketing nowadays requires some level of data manipulation; ChatGPT can be told in layman's terms to generate Visual Basic for Applications (VBA) code that automates complex functions within Excel, saving plenty of time for repetitive actions like sorting or searching and replacing.
We’re already seeing AI step in to deal with simple queries on many customer support calls, escalating only those issues that require a human touch. Whether it’s chatbots or automated phone responses, AI is showing up strongly.
Even in the case of assisting a human customer support agent, AI can help walk an agent through diagnosing problems by asking one question at a time, and ruling out possible solutions until getting to the right one.
The future of AI in the workplace
What’s the future look like for AI?
Personalized AI wearables
Wearable tech, from Google glass to chest-mounted translator pieces, could very well be the future of tech. By intaking data from around a user, comparing data with stored user data, and then absorbing outside sources of data through an Internet uplink, tech can offer on-demand info immediately, doing everything from offering nutritional facts about food and allergies to translating audio or media from another language immediately upon receiving it for the user.
That’s exactly what this TED talk highlights!
In a way, some of these things are already here, in the form of fitness watches from Apple to Fitbit, measuring bodily functions and giving data on demand.
Additionally, devices with haptic responses (touch-oriented wearables, like gloves) can give real time feedback about actions being done by workers — cool, but a bit creepy too.
Autonomously driving AI are a bit in the future — but they’re not that far out. We’re already seeing the first FDA-approved flying car this year, even if it costs $300,000.
A significant portion of the United States relies on transportation, and it’s one of the next frontiers for labor to go. Trucking requires long hours, stressful situations, and plenty of physical strain. The supply chain in the United States totals 2 trillion dollars a year, and trucking is a vital part of it - safety and speed are paramount concerns in an industry that’s hard on its workers, and constantly facing labor shortages.
AI-assisted remote-controlled trucks might be more likely to hit the market than fully autonomous, self-driving trucks, according to Avi Geller, CEO of Maven Machines; his company helps streamline and automate many of the processes in the transportation and trucking industry by adding AI-assisted features.
It’s an ideal place to start revolutionizing, but the big issue is, what happens to the truckers who drive the truck? Remote controlled trucks might be the first step in this case.
Creative work and acting: the next frontier
With the onset of ChatGPT and Hollywood, it’s been a tense situation watching all three unions in the movie industry go on strike: actors, directors, and writers. A lot of that has to do with contracts and the future of work in an industry that guards royalties intensely.
For workers who depend on royalties being paid down the line, it’s understandable: with AI generation, it effectively cuts out the actors, writers, and directors by replacing them completely. In fact, movie studios are considering using AI to clone the likenesses of actors, saving them from having to pay huge contracts for the use of said actors.
While it may seem niche, work is a huge part of the psyche for many people: it provides purpose and income. With AI replicating and, in the process of doing so, devaluing work, what does this bode for the rest of us who aren’t A-list Hollywood stars?
Above or below the API line
Another place where AI, in the form of algorithm-driven apps, is already running rampant is in the sharing economy. Everything from Uber/Lyft to Grubhub/Doordash to TaskRabbit/Rover use some form of algorithm matching to get client users to task users.
Venkatesh Rao, at the Ribbon Farm, points out this non-ideal form of AI: the algorithm-based app turns human labor into something dehumanized and removed from other people. Work that previously was hard to come by has become more easily accessible through the app, but the cost savings are often then passed on to the builders of the API — the software engineers and business executives who are above the API line and performing work of designing the system that assigns work, rather than being part of the sharing economy of being assigned work.
The ethical ramifications of dehumanizing work at play is already showing itself in present day conditions, where the practice of tipping for service through a screen insulates consumers from having to directly deal with facing a service worker, and may reduce tips in some cases for a workforce that is already strapped for income.
Apps that allow you to order food or rides has much of the same issues, by insulating consumers from having more direct contact with those who service them.
So what now?
While there’s a lot of excitement around AI and the capabilities it’ll bring, it could also bring about a bit of an AI-driven dystopia:
The impact of AI on labor day to day will depend on a variety of things: laws around how AI is used is one, but also how workforces are organized around AI.
There’s the flip side: we might be able to unlock AI capabilities for everybody. Imagine everyone having access to being able to code, draw, write, and create at a high level. One thing that AI is doing, especially generative AI like ChatGPT and MidJourney, is putting more emphasis on English and writing majors.
Remember those days when people were scoffed at for having English majors? Well, welcome to a world where ideas are the stuff of life again, and execution/implementation is a matter of guiding AI to do the right stuff. In some sense, it democratizes the process of bringing ideas to fruition, and that might not be such a bad future. We might see a future where education focuses on broadening problem-solving skills in a strategic sense, by seeing different pieces of the puzzle, and then applying AI to execute.