The Research Report is Now Dead Too

In partnership with

Welcome to AutomatED: the newsletter on how to teach better with tech.

In each edition, I share what I have learned — and am learning — about AI and tech in the university classroom. What works, what doesn't, and why.

Today, I take a closer look at Google’s new Deep Research tool, which is to the research report what ChatGPT was to the essay.

I also announce the next AutomatED webinar, which will be on January 31 on Zoom and will focus on advanced techniques to build better AI tutors.

🧰 An AI Tool For Your Toolbox:
 Google’s Deep Research

What is it?

As I reported last week, Google just launched Deep Research, an AI research assistant in Gemini Advanced that autonomously explores complex topics and compiles comprehensive reports with source links. As Google describes it:

This week, I explain how it works in greater detail and give my judgment on its relevance to teaching faculty.

How can one get access to it?

First, go to https://gemini.google.com/advanced and sign up for a Google One AI Premium plan. A one month trial is available for first-time users; the rest pay $9.99 for two months and then $19.99 for month three onward.

Once you’re logged in, click the dropdown menu in the upper left and select 1.5 Pro with Deep Research:

How does it work?

It proceeds through 8 phases:

1. Accepting input (a simple chat/search bar; you input a question or direction):

2. Planning (Gemini 1.5 Pro, the LLM, comes up with the best approach to tackle your question; you wait):

3. Accepting revisions to the plan (a dialogue bar, like chatting with Gemini in other contexts; as you can see in the screenshot above, you can click the “Edit plan” button and then tell it what to change in a conversational way, or not)

4. Researching (Google Search combines with Gemini 1.5 Pro in a way that we cannot observe to search for what the plan specifies; you wait)

5. Analyzing (Gemini 1.5 Pro processes all the text and other media it has gathered; you wait)

6. Creating the report (Gemini 1.5 Pro begins to generate text based in what it analyzed and what the plan concerned; you wait)

7. Delivering the report (Gemini 1.5 Pro delivers a report in preview form, with the ability to open it in its entirety in Docs; you click “Open in Docs” or ask for revisions)

8. Accepting requests for revisions to the output (a dialogue bar, like chatting with Gemini in other contexts; you can tell it what to change in a conversational way, or not)

Finally, since Gemini is in Docs for those with Gemini Advanced, you then have another option for further revisions:

9. Accepting requests for revisions, summaries, analyses of the output (a dialogue bar, like chatting with Gemini in Docs; you can tell it what to change in a conversational way, or not)

✨What do I recommend teaching professors do?

I conducted two preliminary tests of Deep Research’s abilities, each focused on a research question that, for different reasons, I have significant familiarity with:

1. Why has the proportion of adjunct faculty increased at colleges and universities in the US since the 1950s? (some screenshots from this test are included above)

2. What are some of the strongest arguments for and against attributing moral responsibility to AI systems, supposing they qualify as superintelligent (as Sam Altman thinks they soon will) or AGI?

First, I will share what Deep Research did in each case.

After that, I will share my recommendations for what teaching professors should do in light of this new tool.

In short, it is a very big deal for educators who assign research, like ChatGPT was for those who assign essays.

📢 January 31 Webinar:
On Building Better AI Tutors

The results of my subscriber surveys in: the next webinar will be on Zoom on Friday, January 31 from 12pm-1:30pm Eastern time, and it will cover building better AI tutors, with a focus on OpenAI’s custom GPTs, Google’s Gems, and Claude’s Projects, as well as more advanced implementations.

If you want to tutor your students with an AI chatbot — simulating a teaching assistant, a mentor, a peer-reviewer, someone who knows the syllabus really well, or … — this is the webinar for you.

✨Premium subscribers get in for free, while the price is $25 for everyone else.

However, the webinar will assume, as a prerequisite, that you have a solid grip on the basics of building custom GPTs.

If you aren’t confident that you do, I recommend that you take our recently released custom GPT course, now live on LearnWorlds for $25. Those who sign up to the webinar will get 20%-off discount code for the course.

In terms of format, the webinar will be roughly 65% presentation, 35% discussion or Q&A. The presentation part of the webinar will be on more advanced topics like:

• strengths and weaknesses of custom GPTs vs Gems vs Projects

• how to address common tutoring pitfalls with them

• when to pivot to creating more bespoke solutions with LLMs’ APIs, like how to link custom GPTs to other LLMs like LearnLM or how to use AI to regulate itself

In the discussion part of the webinar, you can ask questions, hear what others are doing, and get my take on how to tutor students with these amazing tools.

Sign up here: