|
AI's "Deep Research" Feature Is Stunning
April 2025
Okay, let's get this out of the way up front: this is not another column about using artificial intelligence for shopping. But it just happened that my friend Bill emailed me a shopping-related tech question on the very day that I had planned to investigate the new "deep research" feature.
(People email me tech questions thinking I'm an expert. I simply feed their question into AI and send them the answer. Typically, they're very pleased.)
Bill was interested in buying earbuds and spending less than $100. He wanted to use them to take calls and stream music from his smartphone via Bluetooth wireless while canceling out external noise. He also wanted to know if he could use them to cancel noise while he slept. And he was curious if the JBL Tune Buds Noise-Cancelling True-Wireless Earbuds would do the trick.
So I fed his question to Perplexity's new Deep Research feature (www.perplexity.ai). It's free for five queries per day. (For comparison, OpenAI and Gemini Advanced charge $20/month for access to their Deep Research feature.)
I was astonished. Unlike chatbots, it didn't simply do a quick search and then summarize some reviews. Instead it spent several minutes doing an in-depth analysis that took 14 steps and entailed accessing 30 internet sources before producing an impressive 1,125-word report.
Deep Research lets you watch as it does its thinking. It started out by thinking, "I need to search the web to find information on the best noise-cancelling earbuds under $100." It then showed the search terms it used and the internet sources it read. In the second of its 14 steps, it reported several possible choices but then acknowledged that it still needed to determine whether it's safe or comfortable to sleep with noise-cancelling earbuds. And so on for 12 more steps.
Once Deep Research produces its final report, this thinking process collapses into a drop-down menu that you can access later if you're curious.
The final report was well organized and useful. First it discussed the relative performance of the microphones in budget earbuds. That's not a concern when Bill is listening to music streaming from his phone via Bluetooth. But it gets complicated when Bill talks on the phone via the earbuds because the microphone array needs to pick up Bill's voice clearly while reducing background noise. The report explained this and noted the earbuds with the best microphone performance.
Then it discussed active noise cancelling. This is such an amazing technology: it's not just muffling sound but is actually producing sound waves that cancel out the incoming sound waves to dramatically reduce incoming noise. The report described the performance of different models, including earbuds that reduce ambient noise by 75%.
It then identified and described the three top earbuds regarding microphone performance and noise cancelling.
Next it launched into an analysis of the issues of using earbuds for sleep, including earbud geometry and whether earbuds can be comfortable for people who sleep on their sides. And it detailed several considerations, including citing the Journal of the American Medical Association regarding moisture buildup in the ear canal.
The report then outlined critical tradeoffs and final recommendations. And in case you're in the market for inexpensive earbuds, it concluded:
"For Bill's primary needs--calls, music, and situational active noise cancelling--the EarFun Air Pro 4 delivers unparalleled value at $89.99, provided sleep use is occasional. If call clarity is non-negotiable, the OnePlus Buds 3 at $79 are superior, albeit with shorter battery life. The JBL Tune Buds remain a competent middle option at $69.95, though their active noise cancelling lags behind class leaders. Crucially, Bill should avoid continuous overnight use; dedicated sleep earbuds like the Bose Sleepbuds II (exceeding his budget) remain the safer choice for chronic noise issues."
A key idea behind this step-by-step approach is that as Deep Research investigates, it gets a sense for what it still needs to understand. One step of research provides a basis for the next. It refines its output by incorporating new data and adjusting its conclusions.
Deep Research is the next big thing, and it goes well beyond the chain-of-thought reasoning we discussed in an earlier column.
So what did Bill choose? He went with the recommended EarFun Air Pro, but selected the Pro 3 instead of the Pro 4 because he could get that earlier model for only $49.
Cool gadgets, If I didn't wear hearing aids, I'd buy a pair in a minute.
Kudos to Perplexity for offering Deep Research for free. Also free is their option of using R1, a chain-of-thought reasoning tool developed by a Chinese startup. Again, other AI companies charge for this.
R1 is useful for logic-driven answers such as math problems, software coding, and scientific theories. Their Deep Research offering is ideal for complex topics such as medical studies, market analysis--and earbuds.
Maybe give it a try.
© 2025 by Jim Karpen, Ph.D.
E-mail
Jim Karpen
|
