AI supports a person at the computer.
getty
AI-driven tools are on the rise – but they’re still not where we want them. For example, NinjaTech AI now offers a tool called Atlas that can be used to make appointments, conduct web research, and write code. Calendar planning seems to be a common task for these assistants, but I want more.
My day consists of a little scheduling, jumping into Excel to update some spreadsheets with the latest data from census offices, going into QuickBooks to create an invoice, and going to Constant Contact to send out my newsletter. In each of these examples, the things I do are routine for me; What I do every month or every week is something I have done many times. Some of these apps have productivity enhancements that I haven't experienced since I would have to invest a lot of time to become familiar with a particular tool. What I would like is an AI assistant that can watch what I'm doing and say, “Hey, I can do that next time you need it done.”
I shared my frustration with NinjaTech AI CEO Babak Pahlavan. In an email, he described four “parallel waves” required for my type of personal assistant. The first was self-learning. Teaching the AI to win a game is relatively easy due to the clearly defined rules. But finding the rules for human decisions is quite difficult. AI needs to get ideas, try them out and learn from them. By starting small, with narrow applications like calendar management, you build technical skills to – hopefully – create the full-fledged personal assistant.
Autonomy is Pahlavan's second wave, needed for better assistants. A copilot works alongside a person, allowing the human to control them and correct mistakes on the fly. A fully autonomous personal assistant needs to do things—and do them right—when no one is watching. There are no large language models yet.
The third wave, capability convergence, describes the move from a huge set of very narrow tools to a small set of very broad tools. Each tool has its own rules and style. Workers don’t want to learn how to use multiple tools; You want to learn how to use a tool and have that tool do everything.
Finally, Pahlavan believes we need real-time streaming: “…real assistants are always online, ubiquitous, and available across many channels,…available for iMessage, Slack, call, FaceTime, email, etc., just like a human assistant.” In fact, the AI assistant will be more available than a human as long as the AI doesn't need breaks to go to the bathroom, eat, or sleep.
On the optimistic side, Pahlavan believes that progress is happening very quickly and that in a few years NinjaTech's assistants will be self-learning and ubiquitous.
I ran Pahlavan's take on AI-powered assistants past a person who works on the cutting edge of AI (who prefers to remain nameless). He wrote back: “The fundamental problem is that the model makes too many errors. It's just not smart enough right now. I agree that there are other problems (Babak points out some), but they all seem addressable compared to the model, which just gives the wrong answer too often.”
Current AI models could continually observe what I'm doing and step in to take over tasks, but “the model will do stupid things far too often,” my contact wrote. This is the well-known hallucination problem. The models just make things up and sound very believable. Vectara, an AI company, recently evaluated popular AI models and found hallucination rates between 3% and 27%. The problem is not, as is often reported, simply that the AI repeats falsehoods it has found on the Internet. It regularly invents its own falsehoods.
Top minds are working on the problem and will probably find a way to solve it. The solution could prove to be quite expensive, but we don't know yet. I'm sure they'll eventually get accurate results, or pretty close.
The versatile and ubiquitous AI-powered assistant will make everyone who works with a computer more productive, as well as some who don't even use a computer in their work today. For example, a machine operator in a factory might say to the tablet, “I'm shutting down to calibrate cutting head number 3.” And the tablet responds, “We're three hours away from replacing the belt; Do you want to do this at the same time?”
AI will increase the productivity of a wide range of workers, but we are still a few years away from that.
Follow me up Twitter or LinkedIn. Checkout my website.
When I was 16, I decided I wanted to be an economist, but I also started reading my grandmother's used copies of Forbes. After degrees including a Ph.D. After graduating from Duke and three years as a professor, I found my calling in the business world. I started as a corporate economist (PG&E, Nerco, First Interstate Bank) and then moved into the consulting industry, helping business leaders connect the dots between economics and business decisions. I wrote “Businomics: From the Headlines to Your Bottom Line – How to Profit in Every Economic Cycle” to help business leaders and small business owners understand how the economy affects their businesses. Side trips on this journey include co-authoring a high school economics curriculum called “Thinking Economics” and earning the CFA designation (although I am not an active charter holder). I served four governors on the Oregon Council of Economic Advisors and am currently board chair of the Cascade Policy Institute. My friends and fans love their monthly economic charts, a 60-second scan of the economy. The latest edition is always available at https://www.conerlyconsulting.com/writing/newsletter/. Also note the free subscription link on this page.
Read moreRead less
Comments are closed.