Does OpenAI's Deep Research signal the end of human-only scholarship?

Having got my hands on Sam Altman's latest AI offering yesterday, I'm beginning to wonder when research and scholarship that isn't augmented by AI will be seen as an anachronism

Image: Midjourney

This past Sunday, OpenAI launched Deep Research — an extension of its growing platform of AI tools, and one which the company claims is an “agent that can do work for you independently … at the level of a research analyst.”

I got access to the new tool first thing yesterday morning, and immediately put it to work on a project I’ve been meaning to explore for some time: writing a comprehensive framing paper on navigating advanced technology transitions.

I’m not quite sure what I was expecting, but I didn’t anticipate being impressed as much as I was.

I’m well aware of the debates and discussions around whether current advances in AI are substantial, or merely smoke and mirrors hype. But even given the questions and limitations here, I find myself beginning to question the value of human-only scholarship in the emerging age of AI. And my experiences with Deep Research have only enhanced this.

Deep Research pulls together a number of features from emerging AI platforms to create what is in essence a “research agent” that can iteratively reason its way through complex questions, while drawing on web-based resources and constantly checking its work. Using it feels like giving a team of some of the best minds around PhD-level questions, and having them come back with PhD-level responses — all within a few hours.

Expect, in some respects, the results far surpass what most PhD students are capable of — not because they lack the intellectual capacity, but because they lack the ability to synthesize ideas thinking and data from a vast array of disciplines and sources, and then crunch them into something that far surpasses the restrictive training and perspective that comes with disciplinary boundaries.

Not surprisingly, Deep Research has its flaws, and is still a few steps away from where I’d like an AI research agent to be. But I suspect that reasoning research agents like this will eventually make non AI-augmented scholarship and research look intellectually limited and somewhat quaint.

But back to my first experiences with Deep Research, and what I learned along the way:

Yesterday morning I got up at 5:45 AM as usual, emptied the dishwasher, brewed a cup of tea, and by 6:00 AM I was starting to engage with Deep Research.

I’d spent some time the previous day challenging OpenAI’s o1-pro reasoning model to develop and write a series of seven papers on foundational thinking around navigating advanced technology transitions. The process was much faster than what I could have achieved my own (by a matter of months). But I still found myself feeling that there was something lacking in the depth of simulated thinking and scholarship that the model was showing.

I still need to go through and evaluate/refine those seven papers. But the launch of Deep Research has left me feeling that pursuing them any further would be a waste of my time — such is the rate of AI progress at the moment.

I’ve included my initiating conversation with Deep Research at the end of this article below so you can see how things played out. The process is important as it ended up strongly influencing the final paper that Deep Research produced. It also shows quite nicely the back and forth that the AI model initiated with me to ensure it fully understood what I was looking for.

I chatted with Deep Research about the research question and the framing for it while I ate breakfast. And then around 6:20 AM the AI decided it was ready to go, and let me know “I'll get back to you once I have a solid draft ready for review!”

I had a dentist’s appointment that morning, and so I left Deep Research to do it’s thing, while checking in on it occasionally (you can ask it for progress updates).

Clearly my request was a tough one, because the hours went by and nothing appeared.

Around 9 hours after Deep Research started its assignment (yes, you read that right) I asked how things were going. To give you a sense of what was happening under the bonnet, this is what I got back:

I’d estimate that I’m now at around 75% completion overall. Here’s a quick updated breakdown:

• Research & Reading: ~90% done. I’ve identified and incorporated most key sources and cutting-edge perspectives, though I’m still looking into a few niche areas.

• Outline & Structure: ~95% done. The overall framework and sub-sections are pretty much set, and I’m confident in the flow and organization.

• Drafting (Main Text): ~65% done. Several sections are now in near-final draft form, while others are still partially written but have the main arguments and citations laid out.

• Scenario Development: ~55% done. I’ve fleshed out the outlines of speculative scenarios and started integrating them into the main text. Further refinement is needed to ensure they both illustrate and deepen the central themes.

• References & Citations: ~75% done. I’m maintaining a systematic reference list and polishing in-text citations for consistency, but I still have a bit to do here—especially as new references surface during final writing.

• Final Coherence & Polishing: ~30% done. This covers stylistic refinement, smoothing out transitions between sections, and ensuring the text maintains both scholarly rigor and accessibility.

I’ll continue stitching everything together and refining the prose. Let me know if you have any questions or if you’d like to see any excerpts before the next check-in!

I was loving this — while I was off doing other stuff I had a talented polymathic AI doing some intellectual heavy lifting for me at hundreds of times the speed of a human research assistant.

In the end Deep Research took a little over 12 hours to come back with its final paper. It was a lot longer than I expected, most likely reflecting the complexity of the question I asked. But it also felt it appropriate that the paper wasn’t produced in a few minutes, as the guiding question was far too profound and important for a response that felt trite and formulaic.

So how did Deep Research do?¹

Brilliantly frustrating about sums it up!

After 12 hours, and on asking for another update, I was shown a draft that was supposedly 90% complete. It looked pretty good — the style was a little wooden, but it had a breadth of perspective and framing that were impressive, and plenty of in-text citations.

Then, within a few minutes, Deep Research showed me the final paper.

The prose were vastly improved on the preceding draft. It read very well — although on a deeper read the style is somewhat monotonic through the paper, and lacks the subtle nuances and modulations of expression that a good human writer would insert.

The narrative arc is very good, as is the way that Deep Research integrates thinking and ideas across a vast span of domains to come up with insights that are illuminating and thought provoking.

I set this task in part because navigating advanced technology transitions is not a widely recognized area of scholarship, and I was interested in how the AI would cope with this — especially given my charge to draw on as many areas of thinking and understanding as possible.

I am, however, familiar with many of the disciplinary transcending ideas and framings that inform thinking around technology transitions, and part of this exercise was to see how much Deep Research was able to replicate my own thinking — or to challenge it.

I was impressed. I may have hinted at my own ideas too much in the prompt (you can check this by comparing the conversation with Deep Research and the subsequent paper below). But I was surprised that so many of the connections and ideas in the paper match my own thinking and those of colleagues I work with. What surprised me most here was the focus on responsible innovation, the need to draw on non-technical areas of knowledge and understanding that Deep Research emphasized.

The paper is, perhaps not surprisingly, way too brief in some areas, and lacking in others (I would have liked to see more thinking around what makes this point in human history unique, and some of the more novel challenges that arise as a result). It frustrated me at times that Deep Research seemed to bend over backward to incorporate suggested areas and leads from our initial discussion rather than “thinking” for itself a little more. And overall the paper was not as generative and insightful as a leading transdisciplinary and polymathic scholar would most likely be.

But such scholars are rare, and I know of none who could produce such a solid framing paper around advanced technology transitions in just a few hours.

I’d go so far as to say that, for all it’s flaws, Deep Research’s paper is worth reading by anyone interested in successfully navigating advanced technology transitions — no matter what your area of research, scholarship of expertise is.

But, as I said, it is also flawed — as is the process of using Deep Research.

I mentioned that the 90% draft had in-text citations. The final paper doesn’t — although it does have a bibliography.

This is a big deal as there’s no way to validate claims in the text without painstaking research to match the ideas to the references.

I tried to get Deep Research to revise the paper by adding in-text citations. And this is where the process fell apart.

I’m not sure, but I think that once the final paper had been produced, my AI research assistant gave up the ghost and “died.” My wife thinks that I overworked it! Either way, the responses I go after the final paper seemed like I was back to using the regular version of OpenAI’s AI, without the brilliance of Deep Research.

Subsequent iterations of the paper seemed to be based on the previous conversation, but with no further research involved at all. They were produced rapidly, and in each case were substantially worse than what came before.

I actually felt bereft — I had lost a wonderful synthetic research partner that, even though the relationship was only 12 hours old, had still produced more than I could have achieved alone in months. And now it was gone.

I’m sure that there’s an easy fix to this. But I was left with a great paper that was flawed and unfixable by teh AI that had helped create it.²

But that aside, Deep Research is still pretty damn amazing. To put things into context, it would have taken me several months to create a paper that was even close to what Deep Research generated. And when you consider what I could achieve — in a matter of days — by building on this start, it’s hard to imagine serious scholarship without tools like this surviving in the future.

Unless, that is, the future value of the scholarship lies primarily with the process of creation rather than the relevance and impact of what is created.

To be clear, the technology as it currently stands does not eliminate people as it’s only as good as the directions its given, the sources it has access to, and the subsequent human assessment of what it produces.

But as a scholarship and research accelerator, it’s a potential game changer.

As a rather timely coincidence, today’s episode of the podcast Modem Futura features a conversation with Western University’s Chief AI Officer Mark Daley. We recorded last week — before Deep Research was generally available.

During our conversation I asked Mark whether every leading research institution around the world needs to be using reasoning AI to stay ahead of the game.

Mark’s response was “No, you have to use tools like this to stay in the game.”

Since we recorded this conversation last week, the goalposts have moved yet again. We’re now heading for a time when you have to be using AI tools like Deep Research to even get into the research and scholarship game — never mind stay in the game.

And if you find that a little disconcerting, you probably should.

---

The final paper

You can download the full paper generated by Deep research below. If you’re at all interested in any aspects of how we might start to think about navigating advanced technology transitions in new ways, I’d highly recommend reading it.

And if you’re interested to see how Deep Research threads together thinking across a wide range of domains, I’d definitely recommend downloading and reading it.

As I say above, it is flawed. Interestingly, most of the works cited are valid — at least as far as I’ve been able to tell. Where they didn’t seem correct I’ve flagged this in the paper. But we’ve come along way from hallucinated sources and references!

I’d be very interested in your thoughts on this — please do use the comments, but do read the paper and the process first before providing opinions.

NAVIGATING ADVANCED TECHNOLOGY TRANSITIONS:

Toward New Frameworks for Human Flourishing in an Era of Radical Convergence

ABSTRACT

Human societies have been shaped time and again by profound technological revolutions. Yet as we enter an era where convergent and synergistic innovations in artificial intelligence, neurotechnology, synthetic biology, nanoscale science, climate engineering, and other emerging fields overlap and intersect, we confront shifts of an unprecedented nature and scale. These “advanced technology transitions” are marked by a deepening complexity that challenges conventional assumptions about governance, ethical norms, and even what it means to be human. This paper develops a transdisciplinary exploration of how we might navigate these profound technological transformations in ways that safeguard and enhance human flourishing, while remaining humble about the limits of our predictive and managerial capacities.

The paper begins by grounding the discussion in historical analogies, referencing how societies adapted to transformative inventions such as the printing press, the steam engine, and electricity. It then highlights how the present moment may differ—particularly given the speed, global scale, and deep entanglements among technologies like AI, biotechnology, and advanced materials science. Moving through philosophical considerations of identity and value, governance challenges, sociotechnical complexities, and the ethical frameworks that might guide us, the paper examines why our usual models of decision-making often fail in the face of radically unpredictable change. Speculative but plausible scenarios illustrate potential trajectories for human-technology co-evolution, ranging from “neural economies” to AI-assisted ecological regeneration. While these scenarios are neither definitive nor predictive, they underscore the intricate interplay between technological development, human aspirations, and the fragile systems—biological, social, cultural—on which our flourishing depends.

Ultimately, the paper proposes that navigating advanced technology transitions will require iterative learning, anticipatory and inclusive governance, and a renewed focus on what human flourishing can mean in a world where the line between biological life and technology becomes fluid. It calls for humility, open-ended inquiry, and the willingness to question foundational assumptions about progress, identity, and value. The stakes are high: at risk is not merely economic well-being or national competitiveness, but the very contours of human life in generations to come.

Navigating Advanced Technology Transitions Openai Deep Research 2 3 25

251KB ∙ PDF file

Download

Download

---

Full conversation with Deep Research

This is the complete conversation I had with Deep Research before it started to do its thing. I checked in along the way and made a couple of minor suggestions, but pretty much all of the resulting paper came from this initial back and forth:

Andrew: I'd like you to draft out a thought piece paper on the concept of navigating advanced technology transitions. This is a new and emerging area, and so you won't find much that is directly under this title. Rather, you will need to look for inference from a wide range of fields as to how we might need to think differently about ensuring human flourishing in the face of profoundly transformative technologies in the face of multiple factors that extend to environmental factors, population, resources, geopolitics, human behavior, and how advanced technologies are perturbing the local, national and global landscape in terms of shutting down and opening up possibilities. The paper will need to take account of a wide range of advanced science and technology -- including areas that are not obvious. And it will need to be generative -- developing new ideas on the assumption that conventional ways of thinking, understanding the world, and making decisions, may not work in the future. It should also integrate thinking about technology and mechanisms that ifluence [sic] human flourishing, including governance and decision making. And it should consider the nature of value within a human context, even down to what it means to be human

Deep Research: This is a fascinating and complex topic. To ensure the thought piece is comprehensive and aligns with your vision, could you clarify the following details:

1. Scope of Technologies: Are there specific technological domains you want to highlight (e.g., AI, biotech, nanotech, quantum computing, climate engineering), or should I cast a wide net?

2. Framing the Transition: Should the paper focus on historical technological transitions as reference points, or is it purely forward-looking?

3. Perspective on Human Flourishing: Do you have a specific framework or philosophical lens you’d like to integrate (e.g., Aristotelian flourishing, capability approach, transhumanism)?

4. Governance and Decision-Making: Are there particular governance models or theories (e.g., democratic, technocratic, decentralized) that should be explored?

5. Interdisciplinary Integration: Should I focus on synthesizing existing interdisciplinary insights, or emphasize generating novel, speculative ideas?

6. Tone and Audience: Is this intended for an academic, policy, business, or general readership?

7. Desired Structure: Would you prefer a structured approach (sections like Introduction, Challenges, Opportunities, Frameworks, Conclusion) or a more fluid, exploratory narrative?

Let me know how you'd like to shape this, and I’ll begin drafting the thought piece

Andrew
1. No - the idea of advanced technology transitions is that we need new models, frameworks etc to understand how to navigate transformative, convergent and synergistic technologies. There are some big trends at the moment that I think will be useful to frame thinking, but there are also hidden trends that are equally as important. I don't want you to just stick to these, but as a starting point consider AI, neurotechnologies, biology-based technologies (including gene editing, synthetic biology, the work of people like Michael Levin, biohybrids), automation, robotics, nanoscale science and technology, and synergistic convergences between them. 2. I think the paper must look at the historical context and what we can learn from it. But t also needs to tackle what is unique about this point in human history and whether this justifies new thinking. You need to be historically grounded but generative in terms of forward thinking. You also need to draw on many lines of thought and reasoning across disciplines 3. I would like you to be intellectually critical here and develop your own ideas from what you lean and from first principles 4. I want you to be intellectually generative here as wekk 5. both 6. This is intended for educated and informed readers -- including academics 7. Yes - structured and deep -- make it long

Deep Research
Thank you for the detailed guidance! One last clarification: Do you have a preferred length for the thought piece (e.g., word count or page count)? Additionally, should I incorporate references and citations to existing academic and policy literature, or should it be more conceptual and free-flowing without extensive citation?

Andrew
Yes, please. I'd like you to approach this as a polymath transdisciplinary scholar -- be conservative and cautious in your more radical conclusions, perspectives and ideas, and back everything up with citations, but at the same time look for those insights that are new and potentially transformative -- this is where the generative part comes in. And be humble in your writing -- recognize that bold and radical ideas are worth considering, but may need testing or even be wrong. And as for length, this feels like it should be at least 10,000 words long -- maybe more

Deep Research
Thank you for the clarification. I'll ensure the piece is deeply researched, intellectually rigorous, and transdisciplinary, with citations supporting its claims while also being generative in exploring new ideas. Before I begin, do you have any specific academic disciplines, thinkers, or existing frameworks that you think are particularly relevant and should be engaged with? Also, do you have a preferred citation style (e.g., APA, Chicago, Harvard)?

Andrew
I'll leave the thinkers and disciplines to you - I want to be surprised rather than push you in a certain direction. However, as well as thinkers and writers on science and technology, do make sure you explore scholars in philosophy, science and technology studies, responsible innovation, policy and governance, sociology, social sciences, behavioral science, future studies, and scholars that cross over these areas -- all in the context of the intersection between technology, society, and the future

Deep Research
Great, I’ll incorporate a broad range of disciplines and thinkers across science, technology, philosophy, governance, sociology, behavioral science, and future studies. For citations, do you have a preferred citation style (e.g., APA, Chicago, Harvard), or should I default to a standard academic format?

Andrew
Ok, let's use a standard academic format. My rule on formatting and citations is that they need to be consistent and the reader needs enough information to find the source and verify it

Deep Research
Understood! I'll use a standard academic format and ensure consistency in citations, providing enough detail for readers to verify sources.

Before I begin, do you have any specific thematic areas or key questions you’d like the paper to address in depth, beyond what we’ve discussed? For instance, are there particular concerns about governance, ethics, societal adaptation, or human identity that you’d like emphasized? Or would you prefer I identify the most critical themes as I develop the piece?

Andrew
Please follow the evidence, ideas and insights that you uncover -- again, I would like this to lead to ideas and insights that are new to me, even though I know the field very well

Deep Research
Great, I will take a rigorous transdisciplinary approach, synthesizing insights from a broad range of fields and thinkers while ensuring intellectual humility and caution in drawing conclusions. I will conduct extensive research to ground the thought piece in historical context, contemporary challenges, and forward-looking insights, leading to new and potentially transformative ideas on navigating advanced technology transitions.

Before I begin, would you like any specific balance between theoretical discussions and practical implications? Should I also consider speculative but plausible scenarios to illustrate key ideas, or keep the discussion more grounded in established research and frameworks?

Andrew
You can certainly use speculative but plausible scenarios -- again, be cautious and humble here, but anything that helps contextualize and extend thinking is good. If anything, veer toward the intellectual thinking behind new ideas and perspectives here rather than practical implications. However, you will need to explore potential outcomes of certain actions or inactions in order to justify your research -- this should be driven by questions that are profoundly important -- and because of this you should probably end with thinking around next steps and implications

Deep Research
Got it! I'll integrate speculative but plausible scenarios to illustrate key ideas while maintaining a cautious and humble approach. The emphasis will be on intellectual depth and the thinking behind new perspectives, rather than practical applications. However, I will explore potential outcomes of actions or inactions, ensuring the research is justified by profoundly important questions.

For the conclusion, I'll focus on next steps and implications, guiding further research and thought.

Final clarification: Do you want recommendations or action steps for policymakers, technologists, and other stakeholders, or should the piece remain more open-ended, leaving room for interpretation?

Andrew
No recommendations at this point -- remember the humility bit. I don't think you'll have enough for recommendations. But I do think you should be able to explore possible next steps and possible consequences of following certain pathways

Deep Research
Understood! I'll avoid making recommendations but will explore possible next steps and the consequences of different pathways. Now, before starting, do you have any preferences regarding the structure of citations (e.g., footnotes vs. in-text citations) and the reference format (bibliography at the end, annotated references, etc.)? Also, would you like a summary or abstract at the beginning to provide an overview of the key themes? Let me know so I can ensure the paper aligns with your expectations.

Andrew
Let's use in-text citations and a reference list at the end. And an abstract would be great

Deep Research
Understood! I'll use in-text citations and a reference list at the end, along with an abstract at the beginning. Before I begin, is there anything else you'd like to specify regarding formatting, such as section headings, use of figures/tables, or any particular stylistic preferences? Let me know if there's anything else to ensure the paper meets your expectations.

Andrew
You should definitely use section headings. Use figures and tables where they support the ideas and arguments you develop. On the style front I do not like bullet points or short sections -- prose that develop ideas are preferred. And if this runs to 20,000 words and more that's fine -- there's a lot that needs to be explored here

Deep Research
Understood! I'll ensure the paper is structured with section headings and includes figures and tables where they support key ideas. The writing style will focus on well-developed prose rather than bullet points or short sections. Length is flexible, and I'll allow the depth of exploration to determine the final word count, potentially exceeding 20,000 words if needed.

Before I begin, do you have any preferences regarding the tone—should it be more formal and academic, or would you like it to have a more exploratory and discursive tone while maintaining rigor? Also, should I assume the reader has a high level of familiarity with advanced technologies, or should there be some explanatory context for less specialized audiences?

Andrew
Imagine the tone of a book that has been written by a leading scholar that is accessible to everyday readers, but is also scholarly. There are many scholars who write in this way for a public audience, so you should be in good company

Deep Research
Got it! I'll craft the paper with a tone similar to that of a leading scholar writing for a public audience—scholarly yet accessible to informed everyday readers. This will balance intellectual rigor with clarity and engagement.

I'll now begin the research and drafting process. Expect a deep, structured, and interdisciplinary exploration of navigating advanced technology transitions, integrating historical insights, transdisciplinary scholarship, plausible speculative scenarios, and forward-thinking perspectives on the future. I'll also ensure all claims are backed by citations and the work is framed with humility, recognizing the complexity of the subject.

I'll get back to you once I have a solid draft ready for review!

---

Updates

Feb 4

Playing around more with Deep Research it seems that there are inconsistencies in execution on the web based interface. You can invoke it when running the 4o and o1 pro models, but the behavior is very different — and I get the impression that users are just supposed to be using 4o at the moment.

With o1 pro ChatGPT has a habit of seeming to stall, and often not responding to questions after this. But then — and somewhat randomly — it will give progress reports if asked. But there is no progress bar or sidebar, as touted by OpenAI.

With 4o you do sometimes get both a progress bar and a sidebar showing you progress. But not always. It’s almost as if OpenAI are either testing different configurations on the fly, or the interface is still somewhat flakey.

The quality of the analyses also seems to be very variable. My 12 hour marathon above was great. But i’ve had analyses running Deep Research today from 4o that are just mediocre — and with plenty of false references.

Seems that things are still being worked out.

1. I have to comment at this point that “Deep Research” is just too close to the computer “Deep Thought” from Douglas Adams’ The Hitch Hikers’ Guide to the future — to the extent that I half expected it to come back and tell me that the answer was simple, but I needed a more powerful AI to work out what the question is!

2. I’d also add here that one of the frustrations of working with a tool like ChatGPT on a paper like this is that every new draft is a rewrite — and often a substantially different one. If you see something that you like that just needs a few tweaks to move it from good to great, good luck!