Google is expanding Gemini, its chatbot platform, to incorporate the power to “justify” a research problem and produce a comprehensive report.
The updated Gemini offers a feature called “Deep Research,” which Google says uses “advanced reasoning” and “long-context capabilities” to supply research reports. The briefings are presented within the Gemini apps and will be exported to Google Docs for further editing.
Currently, Deep Research is on the market exclusively on Gemini Advanced, a more sophisticated version of Gemini that sits behind the Google One AI Premium plan and costs $20 per 30 days.
Dive deep
According to Google, deep research can analyze information relevant to a search query from across the net on behalf of a user, acting as a sort of research assistant. The results are summarized within the briefing and supplied with links to the unique source material.
This is how it really works:
- A user writes a matter;
- Deep Research creates a “multi-stage research plan” that the user can either revise or approve;
- Once the user agrees, Deep Research refines its evaluation over the course of a couple of minutes – searching, storing potentially interesting information, after which starting a brand new search based on the knowledge
learned; learned - The process is repeated several times and once complete, Deep Research produces a report with the important thing findings.
Deep Research is initially only available in English on the desktop and mobile web. Users can access it by choosing “Gemini 1.5 Pro with Deep Research” from the model’s drop-down menu. Google says Gemini mobile apps shall be available in early 2025.
Google's recent Deep Research feature appears to be one of the appropriately “Google” applications of AI so far, and it's quite impressive.
I've had access for a while and it provides excellent initial reports on almost every topic. The paywalls surrounding academic sources impose some limits. pic.twitter.com/dwSqr6aKGZ
– Ethan Mollick (@emollick) December 11, 2024
“We've developed a brand new agent system that leverages Google's expertise find relevant information on the internet to guide Gemini's browsing and research,” David Citron, product director for the Gemini apps, wrote in a blog post to TechCrunch . “Deep research saves you hours of time.”
But the feature raises all types of thorny ethical questions.
Possible damage
Aside from the incontrovertible fact that all AI makes mistakes and hallucinates (anyone remember sticky pizza?), technology like deep research could have serious consequences for education.
In a current one op ed In the New York Times, Jessica Grose wrote about how students are increasingly counting on generative AI to outsource brainstorming and writing. She said these students are liable to losing critical considering skills and overcoming frustration with tasks that can’t be easily accomplished.
At least one study has linked heavy ChatGPT use amongst students to higher levels of procrastination, memory loss, and lower grade point averages.
Deep Research also threatens to cause financial harm to the publishers from which it obtains its information. By scraping information from web sites and compiling it into briefings, deep research could deprive those web sites of invaluable promoting revenue.
The impact on publishers of AI Summaries, the AI-generated summaries that Google provides for specific Google searches, could also be a sign of what's to come back when deep research takes off.
After According to 1 source, publishers have seen a 5% to 10% decline in search traffic since launching AI Overviews earlier this yr. On the revenue side, an authority cited by the New York Post appreciated that AI-generated overviews could lead to losses of greater than $2 billion for publishers.
Google often says it really works closely with publisher partners, respects paywalls and allows sites to dam its AI scraping on the domain level. But outlets often face an issue dilemma: Allow scraping or lose visibility in Google Search (not less than for now).
Google claims deep research can “connect users to relevant web sites they won’t otherwise have found, in order that they can dive deeper and learn more.” We'll need to see whether the feature delivers on that promise of discoverability or diverts attention from the broader web.
Gemini 2.0 Flash
Deep research isn't the one recent capability Gemini offers.
Starting today, each free and paid Gemini users will gain access to Gemini 2.0 Flash, Google's latest flagship AI model. More specifically, it's an experimental version of two.0 Flash optimized for chat – the complete version shall be released in January.
Google says Flash 2.0, which will be chosen from the Gemini model drop-down list on the desktop and mobile web (but not yet within the mobile apps), should offer higher performance on a variety of tasks – and faster answers.
Still, the corporate warns that some Gemini features “is not going to be compatible with the model in its experimental state.” It wasn't stated exactly what features there are.
