AI Talk: So... Who’s Doing the Research Now?

Tis is the golden era for research! All the world at ones fingertips. Lets explore what the world of AI is now providing in terms of assistance for your favorite research project.

An image of a robot secretary ready for any assignment - Created by Gemini

Google Notebook LM

This is one of my favorite tools. It shot into prominence last year with its ability to create a podcast on just about any paper you upload for it to analyze. The podcasts all sound somewhat the same, but they’re quite thorough and engaging. However, they avoid criticizing the paper—an academic no-no in some circles.

Fast forward a year, and this tool is evolving. Now you can interrupt the conversation and ask your own questions, and the AI agents will respond and continue the discussion. It also now supports several new features that make it especially useful for students—their main target audience. These include a study guide, an FAQ, a briefing document, and a MindMap of key concepts extracted from the papers in memory.

Deep Research

All the major LLMs now offer a “deep research” option. What does it actually do? I tried a simple test: chart the evolution of Generative AI from the Transformer to today’s systems. I used the same prompt on ChatGPT o3, Gemini 2.5 Pro, and Claude Opus 4.

You can even click “show thinking” and—boy, do they think! In this case, it means scouring hundreds of websites while you sit back and marvel. They all produce extensive reports, with references. You could use them to write book chapters! (And I suspect many already are—hence the deluge of mediocre AI books floating around.)

Gemini, for example, provides two sets of references: one it used to write the report, and another it claims it read but didn’t cite. I liked that I could generate PowerPoint decks explaining the evolution of Gen AI. I’m now all set to wax eloquent on the subject while my AI assistants do the legwork—leaving me free to pretend I’m a CS professor. Just remember to double-check the links.

Ai2 Scholar QA

I have been using semantic scholar for over a decade—and I love that it lets me track new research in topics I think I’m interested in. (Tracking doesn’t mean I actually read them. Let’s be honest—the volume is bonkers.)

Semantic Scholar now has built-in AI that highlights key elements in a paper to make skimming easier. This tool is from the Allen Institute for AI (AI2). Earlier this year, Ai2 released "Ai2 Scholar QA.", which I also tested using the same Gen AI evolution prompt.

Unlike the deep research tools that crawl the web, this one digs through a curated corpus of over 8 million open-access papers to generate its report. A godsend for anyone writing background sections in papers—and I must say, quite impressive.

Khan Academy

I’ve been a huge fan of Khan Academy for decades. Sal Khan, the founder, is a consummate teacher. His hand-drawn figures and easygoing style made complex topics surprisingly digestible.

Fast forward two decades and Khan Academy is fully immersed in the AI world. They’ve integrated Gen AI into a tool called Khanmigo. While the AI2 and deep research tools are geared toward adults, Khanmigo is designed for the rest of us—especially K–12. It offers a slew of tools to help teachers, parents, and kids alike. Microsoft is a key supporter, and partnerships span CMU, the entire state of New Hampshire, and even schools in Telangana, India!

The Specialist Squad: Not Your Average Chatbot

Sure, your everyday chatbot can schedule meetings or fake small talk at dinner parties, but these AI assistants are playing a whole different game. These are the nerds of the AI world—happily buried in proteins, proofs, and petabytes of data. And they’re quietly powering the next scientific revolution.

Let’s start with biology, where DeepMind’s AlphaFold 3 and the open-source RoseTTAFold are folding proteins with an accuracy that would make origami masters weep. This isn’t just academic—it’s reshaping drug discovery. Meanwhile, Isomorphic Labs is trying to monetize the magic, and Google’s experimental AMIE Medical is learning to play doctor—and recently gained the ability to see (yes, vision features are now enabled).

Over in chemistry, IBM’s RoboRXN is predicting chemical reactions like a caffeinated oracle, while Microsoft’s AI for Scientific Discovery is on a quest to uncover new materials—possibly even the next unobtanium.

Math and computer science aren’t being left out either. Google’s AlphaProof tackles math proofs with the flair of a tenured professor, and AlphaEvolve (still in the lab) is trying to out-code human programmers - an eternal quest for a better algorithms. Sleep is optional for AIs, after all.

Bottom line? 

The lonely scientist is becoming a thing of the past. The new research dream team includes AI specialists who don’t take lunch breaks and never forget a citation and throw in a few which doesn't exist!

Acknowledgements

My AI assistants - ChatGPT, Gemini, Claude.