If Everyone Could Afford an Assistant, What Might the World Look Like?
(This was originally published on Medium in March 2024)
Few months back I was exposed to an app called Pi¹. It is designed to be your ideation assistant. Intrigued, I downloaded it and since then, have watched the app options get richer with every update. In somewhat fun and endearing ways, you can simply vent to Pi, discuss your cognitive weak spots, or brainstorm with it on your daily productivity. And it is all free (at least for now).
It got me thinking about what the future holds. With tools like Generative AI becoming increasingly general-purpose technology, what roles will they assume in our lives? One could start by asking the question, “What has history taught us about the progression of general-purpose technology?”
Most technologies don’t start off as general-purpose. Initially, it is a niche idea which, over time, gains adoption in multiple industries and use cases. As time passes, it slowly gets intertwined with more aspects of society, industry, and our daily life. At that stage, it somehow stops being distinguishable as a technology per se. One could see that with electricity, the internet, cars, or television. We take these for granted even though each of these ideas is quite remarkable in what it has made possible.
Let’s take the internet — some of us remember the white noise of a 56kbps dial-up modem connecting to the information superhighway in the early 2000s. It was painfully slow and erratic. Nowadays, a 60Mbps (or 60,000kbps) connection is very common. That’s over a thousand times faster. So now, simply unlock your phone, a couple of clicks, and you have access to almost any information or entertainment, anytime you want it!
AI seems poised at an early stage of such a cycle. One could argue it’s all hype and may not achieve its potential. Some part of the story does feel overblown. I have witnessed the “irrational exuberance” of the dot com boom of the late 1990s. There are parallels one could draw from it. Several companies changed their names to .com (like they do with .ai today), even though they had nothing to do with the internet. There were talks about everything being delivered on the internet, hence the addition of .com to your company name made sense. The stock prices of any company with a .com extension went through the roof (remember Pets.com and Webvan?). The general sentiment was people ‘not on this bus’ will miss out on the biggest financial upside the world has ever seen.
We all now know how that panned out. The bubble did burst, but the internet still became pervasive². Most things do get delivered online today, many industries got reshaped by the internet and technologies like cloud that came after it.
AI will most likely follow a similar path. The reason for that is threefold. One, Moore’s Law has ensured enough headroom in the compute infrastructure to crunch enormous amounts of data. Second, research in machine learning has given us techniques capable of training this data. Third, the network and storage capabilities have grown equally fast to retain or transfer incredible quantities of data at unimaginable speeds. As the past has demonstrated, these three components will only get better and cheaper with time. Such is the human capacity for relentless innovation.
Out of all the use cases that AI may enable, the one that seems most probable to me in the medium term is one of an assistant to every human being. The proverbial Watson to Mr. Holmes. The key difference though is this assistant never tires, is available 24x7, and savvy in multiple fields. In fact, its speed of thoughtful responses could stun Mr. Holmes himself. The ever-reducing cost curves of hardware technology and ever-improving Generative AI (or follow-on) tools will make these assistants pervasive across domains in my view.
The simplest analogy one can think of is that of a calculator for anyone who needed to manipulate large volumes of numeric data before the advent of PCs or smartphones. It is a tool that was readily available, extremely affordable, versatile, and given its size, very accessible too. It dramatically increased your productivity. I remember as an engineering student one was allowed to use a scientific calculator for some subjects.
With assistants being available at our beck and call, how they will manifest themselves in different industries was my next question. At the highest level, any area that has a plethora of theoretical possibilities impossible for humans to fathom or analyze manually are the typical ones where these assistants could make the most impact. With that criterion in mind, here were some examples I could think of:
Software development: Generative AI powered Copilot is already making software coding easier. GitHub data shows almost 50% of the code on average is being written by Copilot today³. Over time, it is very likely that all repetitive tasks like testing or QA may get delegated to such tools.
Media and Entertainment: Until very recently prompts could only generate text or images. But now videos have been made possible. Instead of simply writing a movie script, one could create a short snippet to show the producer. You could choose any star you prefer and all it will take is a precise prompt to your assistant. There are tools available today that can create a podcast for you on a topic of your choice. I believe all creative fields will witness a rising tide of ‘synthetic ideas’.
Professional Sport: A coach helps an athlete excel by removing their blind spots or improving their technique. An intelligent assistant could watch the video of your recent games and analyze the gaps. It could uncover patterns of strategic lapses, biomechanics of repetitive motion (potential cause of injury), advantage over specific competitors and much more. It could prove pivotal for those athletes who want to raise their trajectory of performance.
Healthcare and Drug Research: Doctors will rely on an assistant up to date with all recent medical industry data to accurately assess the patient condition compared with those with similar symptoms. They could debate the treatment options with the assistant too. Who knows, if genetic data for all patients is made available in a protected database, such an assistant could likely recommend precise medication for any patient⁴.
Chemistry in Material Design: Any design, whether it is a 3D model of an EV, layout of a space capsule, molecular design of a new compound or a unique food recipe, can be generated in quick time. The assistant could help brainstorm or refine your product idea keeping novelty and feasibility in mind. It could deliver sustainable solutions to problems humanity has battled for decades. What if a research chemist along with their ‘assistant’ creates a synthetic fuel compound with zero carbon emission? Or a compound luminescent at room temperature that lights up our homes without any energy consumption?
Education: Teachers could lean on their assistant to personalize lesson plans for each student, deliberate on assessments to test the capability levels of students. Remember, these students will have support from their own assistants to respond to those.
Financial Services: Anybody (or ‘anything’ to be precise) who can be implicitly trusted, is market-savvy, understands my portfolio needs, advises me on the trade-offs of various investment options and warns me of the risks ahead of time could be a very valuable resource. An objective assistant without the vested interest in benefiting from your wealth could be vastly more beneficial than manipulative brokers today.
Suffice to say, over time, one can think of an assistant being made possible in every walk of life. What are its implications to us and our society?
Firstly, the baseline for every skill will get raised. The flip side of this argument is those who don’t use the tool, citing purity of ideas or any other argument, will degrade from the mean. Of course, there will be exceptions who are so talented, they won’t need any prop. But for the vast majority, tools will add both creativity and productivity. Anyone could inexpensively create an unlimited set of ideas.
To test this, I asked ChatGPT to describe a scientific calculator. It responded with, “Before computers, engineering graduates relied on compact, sturdy scientific calculators featuring a keypad with functions for arithmetic, trigonometry, and calculus. These calculators had small LCD screens, memory functions, and ran on batteries, providing essential computational power for complex calculations in a portable package.”
I then used this description as a prompt on Microsoft Designer asking for funny images of a scientific calculator. Not the coolest output but that isn’t the point. Just think about all the divergent ideas you could unleash with these tools!
Secondly, the caliber of output of the most talented in any field who make the best use of such tools will get significantly improved. The best will get off the charts better than most of their competitors. Hence, outliers in any domain will be those who are talented but also adept at making the assistant work the hardest for them.
The second-order effects of the above could be increased wage disparity. The most prolific in any field will command a higher premium while the average may just about make ends meet. I find it hard to imagine any scenario where companies don’t choose to maximize productivity or profitability at the expense of reduced manpower.
Assistants spanning every facet of our life also raise questions around data privacy and security. How our health or financial data gets used, as an example, will require a thoughtful approach on topics like ethics and algorithmic bias.
But it could also result in spawning new industries that will absorb the labor not required in the current businesses — a classic argument against the Luddites. However, I don’t believe new industries will absorb talent impacted in traditional jobs. It will be the most skilled who will make that transition.
Proactively addressing such adverse impact won’t be easy. The mitigation may lie in encouraging continuous learning and skill development. Fostering a culture of curiosity, subsidizing education, and creating public recognition systems that excel in learning new skills could be ways to empower citizens. It will create an environment where people feel motivated and supported in their pursuit of learning.
Some, including OpenAI, have spoken about Universal Basic Income (UBI)⁵ as a way to address income disparity. While it looks good in theory, I don’t believe it is practical. It could potentially disincentivize some individuals from pursuing meaningful work or personal growth. Perhaps a combination of UBI and skilling initiatives may strike a better balance instead of simply giving handouts.
Standards help in scaling and safe use of any technology. Automotive and aviation are case studies of standards enabling a global industry yet ensuring safety. Both have transformed transportation with relatively low fatalities.
AI is an area where global standards for ethics and governance will prove vital in my view⁶. Technology like AI is analogous to nuclear fusion in terms of scale of impact. Countries collaborating towards a common cause instead of competing will uplift everyone. But long-term thinking will be required for such collaboration. A safe, high-trust, and ethical environment is one where AI technology can flourish and serve humanity.
That brings me to the final question, “What should I do now to stay relevant in my career?”
Looking at the past can provide us an answer. In any disruptive transition, history shows the people or companies that emerge successful are those who embrace the change and best position themselves regardless of the outcome (which nobody can predict).
As a start, our mindset defines everything in such an environment. One needs to think about AI tools literally as our assistant — something that will augment our work instead of replacing us. Best case, even without a college degree, it could help us become an expert in a domain unknown to us.
Approaching it that way will highlight our blind spots and being very savvy on how to collaborate with assistants to address them. There is no debate that such tools will raise the productivity bar. Those who don’t use it will find it harder to contribute beyond the baseline that a machine can deliver right off the bat.
At the same time, understanding of the fundamentals including human psychology will gain importance. Though Copilot could generate most of the code, you still need the programming skill to assess if that code is doing exactly what it is supposed to in a way you want it to. Tools may generate any art, but you still need to determine which option will best differentiate yourself as an artist. AI may present you with alternatives, but you will still be responsible to judge ‘the right thing to do’.
So, while tools may make us more efficient, it is only our grasp of foundational and soft skills that will raise our effectiveness with them. To that end, a deep understanding of our domain and being human will never go out of fashion, in my opinion.
Next will be being well-versed on tool usage. While the learning options in this space may seem overwhelming, starting small by exploring the basics may not be a bad idea. The pace of change in these tools is extremely high. That also means one can step on the bus any time and not be outdated. The best approach will be to keep ourselves current through online courses, hands-on sessions (if your job demands it), engaging with experts, and staying insatiably curious.
It is easy to get carried away by the daily information frenzy surrounding AI. An approach that has helped me in such times is to step back, slow down, and not panic. Reduced pace brings clarity. The world is rarely moving so fast as what our social or news feed likes us to believe.
While there is no doubt that AI signals significant change, we shouldn’t forget that ultimately any technology is simply a tool. Tools like assistants in the hands of the most curious, talented, trustworthy, and kind individuals could be the difference between being simply good instead of being recognized as the best in any field.
We may be living through the AI hype phase, but this is the opportune time to sharpen your axe. The whetstone is waiting to be prompted.
And am reasonably certain of one thing — there is enough time, the game just got started.
Notes:
1. Forbes article on Pi: Inflection AI, Startup From Ex-DeepMind Leaders, Launches Pi — A Chattier Chatbot (forbes.com)
2. Irrational Exuberance of dot com boom: Turns Out the Dot-Com Bust’s Worst Flops Were Actually Fantastic Ideas | WIRED
3. GitHub CEO interview: GitHub CEO says Copilot will write 80% of code “sooner than later” (freethink.com)
4. Eric Schmidt view on impact of AI in sciences: Eric Schmidt: This is how AI will transform how science gets done | MIT Technology Review
5. Sam Altman’s article on AI: Moore’s Law for Everything (samaltman.com)
6. Global AI governance commentary on Brookings: Should the UN govern global AI? | Brookings