AI Beat: Why does OpenAI need a search engine?

If these reports and rumors are true, OpenAI, the maker of ChatGPT and arguably the company most responsible for leading the generative AI craze in Silicon Valley and globally (in terms of investment dollars and product offerings), is gearing up to launch as early as May 13 A new artificial intelligence search engine is launching on Monday.

But as regular users and paying subscribers of its flagship product, ChatGPT, know, you can already get AI-driven search directly through ChatGPT today.

Just prompt/ask the chatbot (using GPT-4) to search the web for some topic or information, and it will automatically go into Bing Search (from OpenAI supporter and ally Microsoft) and display the information along with links and citations.

---

Coming soon: AI auditing with UiPath

The latest stop on the AI ​​impact journey is coming!

Join us for an exclusive live event in New York on June 5 as we dive into the fascinating world of AI auditing in our latest event with partner UiPath.

Explore strategies for checking AI models for fairness, performance, and ethical compliance.

You don’t want to miss the opportunity to meet other industry leaders. Learn more: https://bit.ly/4b8Psak

---

How is the OpenAI search engine different from ChatGPT + Bing?

How could a separate or unique OpenAI search engine improve this experience enough to warrant investment in developing and launching new products?

Without understanding OpenAI’s plans, it’s difficult to judge or draw any clear conclusions about why a generative AI search experience alone would be more beneficial to users than current offerings, or how it would attract more potential users.

In fact, there are already other AI-powered search engines, namely Perplexity and Google’s Search Generated Experience (SGE), the latter of which is an opt-in ad that users can open when logged into their personalized Google account/profile.

But perhaps it’s the existence of these services that prompted OpenAI to launch a competitor.

Competition ignites the fire

After all, Perplexity has received millions of dollars in funding for its product — which allows users to choose between different open source and proprietary AI models to power their searches — and has also quickly won over enterprises with its private internal data search capabilities. favor.

Since OpenAI is also keen on attracting and winning enterprise customers, it seemed natural for them to offer a search product, as this is what some enterprise customers have been clamoring for – although the search product is already baked into ChatGPT.

ChatGPT for conversation, analysis, and work, and OpenAI search for research?

Maybe ChatGPT is actually so good and all-encompassing that OpenAI sees an opportunity to break it down a bit into individual services and products, such as forward search services and products.

Perhaps the company has done research or believes (correctly or not) that users view ChatGPT as a conversational assistant and content creation tool rather than a research tool, and wants to provide clarity and uniqueness for both use cases The product.

If so, it’s an interesting strategy for the company to bundle its image-generating AI model DALL-E 3 directly into ChatGPT.

Is more just more?

In recent months, however, it’s become clear that OpenAI’s products also embrace the “more is more” philosophy that drives its custom GPT Builder and GPT Store, and its Model Spec release suggests it’s considering allowing its applications Users of the programming interface (API) have even generated “not safe for work” NSFW/explicit content (even pornographic content), which the company has thus far sought to suppress and prevent users from using its products.

OpenAI may take the approach of Apple under Tim Cook – have multiple different products with similar features and use cases, but slightly different configurations for different user groups and different price points.

This approach to business has some advantages: Whatever the needs of individual customers, no matter how unique, the message to them is that “we have a product among many that we hope will work for you.”

But it’s also a dangerous path, as it can quickly overwhelm users with too many choices and paralyze them, slowing down their decision-making process or causing them to abandon the product altogether in favor of simpler, more A cleaner, more “single” product. A “one-size-fits-all” option from competitors.

Regardless, we may soon find out that OpenAI’s vision for search extends beyond the current ChatGPT and Bing integration, and from there we can start using it and analyzing it.

Is it better than Google or Perplexity? What are its advantages and disadvantages compared to other products? Who is it directed at? Why would some use it and not others? What impact, if any, may it have on the search market and the broader market for search resources.

The latter issue is of course a particular concern for publishers like VentureBeat, who rely heavily on Google search recommendations to attract visitors and readers to our articles, and which we can get by showing ads directly to users on our site advertising revenue.

Publishers complain and deal with the devil

Google has come under growing criticism in recent months and years for robbing publishers of ad revenue by keeping visitors on Google search engine results pages (SERPs) and in ways that appear to disadvantage smaller independents. The way publishers adjust their algorithms, these independent publishers are now riding a wave of AI-generated content that in many cases directly plagiarizes their own work (or indirectly benefits from it, as Google’s AI products and larger ones generate Artificial Intelligence Large Language Model Corpus) [LLMS] Trained on a large amount of publicly posted web content).

Would OpenAI’s putative search product face the same criticism for providing users with information pulled from source articles without incentivizing users to actually visit those sources and provide them with traffic and ad impressions?

OpenAI has subtly tried to head off some of the criticism by signing licensing deals directly with publishers including the Associated Press, Politico and Business Insider publisher Axel Springer, and most recently the Financial Times – where it plans to make their content available or where it A summary of the response in ChatGPT along with a link back to the source.

generate generation

Regardless of what happens, though, it’s clear that the era of outbound link ranking web search pioneered by Google over 25 years ago is changing rapidly, and I’d be willing to bet that it’s coming to an end.

Especially at this time, it makes sense. To put it in perspective, 25 years is a long time in technological and even human history—a span of one generation. Now, a new generation is rising: the creative generation, to put it mildly.

Perhaps that’s why OpenAI seems ready to take its next step into search.

In an age where software robots can not only crawl and index the web, but can now read, summarize, translate, analyze, and interpret it, why do we humans still search the old-fashioned way like we did in 1997?

Google’s decline in perceived quality leaves a hole

Furthermore, to be fair, Google’s search experience has been widely blamed in recent years on the company itself, as the top of the SERPs has become increasingly filled with sponsored results placements and information boxes (which it calls “knowledge panels”). Varying accuracy and helpfulness Instead of returning a link to the source of information the user is looking for.

The company has even been sued by the U.S. government on antitrust charges for this tendency (as has Amazon).

Which brings up the real question behind it all: What is search used for? Are people trying to solve problems and are their questions answered as quickly as possible? Or are they looking for potential source material to help them answer the question themselves?

There may be room for multiple approaches. However, in an era shaped by Google’s overwhelming search dominance, which is rapidly being challenged by artificial intelligence, it’s hard to imagine one or a few companies — OpenAI, Google itself, Perplexity, Bing, and a host of other startups — Scenes. Rather than becoming the leader of the next era of search, perhaps the leader of the web itself.

I don’t know if we can answer now why OpenAI thinks it needs a search engine, but going back to the original question, I daresay the answer is: because it can build one, maybe even one better than what we have now. At least, that’s presumably what Altman and his colleagues hope.

That’s all for this week! have a good weekend. TGIF. Get out safely.