The Memo - Special edition - OpenAI deep research - Feb/2025
Example outputs from the new OpenAI deep research model, based on o3
To: US Govt, major govts, Microsoft, Apple, NVIDIA, Alphabet, Amazon, Meta, Tesla, Citi, Tencent, IBM, & 10,000+ more recipients…
From: Dr Alan D. Thompson <LifeArchitect.ai>
Sent: 7/Feb/2025
Subject: The Memo - AI that matters, as it happens, in plain English
AGI: 88%
ASI: 0/50 (no expected movement until post-AGI)OpenAI CEO (2/Feb/2025):
‘[OpenAI deep research] is a system that I think can do a single-digit percentage of all economically valuable tasks in the world.
This is a huge step forward for AI.’
OpenAI ‘deep research’ (lowercase, according to OpenAI) is the latest state-of-the-art model based on the new o3 reasoning model. It was released on 2/Feb/2025, and is currently available to ChatGPT Pro users (US$200/month). Much of the functionality has existed for a while in platforms like Perplexity (Aug/2022), Meta AI Galactica (Nov/2022), and Google Deep Research (Dec/2024). OpenAI said (2/Feb/2025):
Deep research was trained using end-to-end reinforcement learning on hard browsing and reasoning tasks across a range of domains. Through that training, it learned to plan and execute a multi-step trajectory to find the data it needs, backtracking and reacting to real-time information where necessary. The model is also able to browse over user uploaded files, plot and iterate on graphs using the python tool, embed both generated graphs and images from websites in its responses, and cite specific sentences or passages from its sources.
I demonstrate a range of text prompts in some of my areas of interest, as well as some interesting topics from others. Deep research took an average of 11 minutes from prompt to response, and it looks like hallucinations have been practically eliminated. In this edition, I’ve provided a link to the prompt and any clarifying questions, a ‘content rating’, and a download of the final output file as PDF.
At launch, deep research is not fine-tuned to help with the presentation layer (e.g. via LaTeX, templates in Docs or Word), although it’s expected that this would be an easy addition for future versions. For now, I’ve had to manually present the content in various formats for completeness.
Imaginary Friends in Exceptionally Gifted Children (IQ 150+)
Prompt: Write a paper on exceptionally gifted children (IQ 150+) and imaginary friends. Aim for 5-10 pages. (see the full chat)
Quality rating (content): High quality. Decent coverage of the gold-standard academic researchers in the field, with effective sourcing of specialized case studies in this focused area. ★★★★☆
Download (7 pages, formatted as paper using Overleaf/LaTeX):
Banjo Paterson road trip itinerary
Prompt (by Jess the editor): Create a travel itinerary around Australia based on locations mentioned in Banjo Paterson's poems and stories. (see the full chat)
Quality rating (content): High quality. Missing some important stops, but does uncover hidden (nearly Google-proof) events like the ‘Waltzing Matilda Bush Poetry Festival’. ★★★★☆
Download (26 pages, formatted as report using Google Docs):
Microphone use in arenas and stadiums: Technical analysis
Prompt: Write me a paper about microphone use in arenas and stadiums. Focus on male vs female, HPFs, and patterns. Give me the final PDF. (see the full chat)
Quality rating (content): Very limited context, probably due to prompt request for it to be ‘short-ish. 3 pages.’ ★★☆☆☆
Download (2 pages, original PDF format by deep research):
Niche and boutique signature fragrances for men
Prompt: I need a detailed technical report on signature fragrances for men. Do not include any designer options, only niche and boutique. This is for a thesis, so references are important (APA style). (see the full chat)
Quality rating (content): Completely missed several families from the fragrance wheel (fresh, floral) which led it down a limited path. Poor selection of credible sources. ★★☆☆☆
Download (12 pages, formatted as report using Google Docs):
Bitcoin vs. fiat money: Benefits, drawbacks, and what it means for you
Prompt (by Jess the editor): What are the benefits of bitcoin and/or cryptocurrencies, especially in contrast to current fiat monies? (see the full chat)
Quality rating (content): Prompt was ‘targeted to beginner audience,’ so as expected this is a fairly average output, with basic coverage of benefits and pitfalls. ★★★☆☆
Download (36 pages, formatted as eBook using Vellum, links removed for print):
Towable off-grid ablution block options
Prompt (by Jess the editor): Write me a paper about purchasing or making a portable toilet/ablution block for rural property in northern NSW. (see the full chat)
Quality rating (content): Rigorous sources, excellent suggestions to broaden original prompt to include local regulations. Accurate details on relevant models. ★★★★☆
Download (12 pages, formatted as report using Google Docs):
OpenAI o4 and AI Reasoning Models
Prompt: Write a detailed technical report on the upcoming OpenAI o4 reasoning model (lowercase). Use tables and diagrams. You'll have to speculate on o4, but use evidence-based estimates from recent models like DeepSeek-R1 and OpenAI o3. Use OpenAI's previous reports for layout, LaTeX format.