Three Key Takeaways from Meta's Galactica AI

Galactica AI,  general idea is good, navigating complex forms of data is useful.  One of our early AI successes did exactly this for key internal knowledge.  Make it easy to easy to use, and importantly make it easy to validate and update in changing contexts.

Three Takeaways from Meta's Galactica AI

By TechTalks, November 23, 2022

We should look at Galactica's initial failure as another scientific experiment...and, every failed experiment brings us one step closer to success.

Meta's most recent large language model (LLM) release, Galactica, was supposed to help scientists navigate tons of published scientific information. Its developers presented it as being able to find citations, summarize academic literature, solve math problems, and perform other tasks that help scientists in research and writing papers.

However, three days after the release, Meta had to shut down the online demo following a deluge of criticism by scientists and tech media about the model's incorrect and biased output. While Galactica was obviously not a success, its short history provides us with some useful lessons about LLMs and the future of AI research.

From TechTalks   View Full Article    

About Xerox Parc

 Worked some with and met at PARC early on.    As a group that wanted to create the architecture of information, always wondered what happened to them.   They did own the computing infrastructure.     Were they really scalped by Gates and Jobs?  What do they do now?  Here a good telling of history.

XEROX PARC’S ENGINEERS ON HOW THEY INVENTED THE FUTURE—AND HOW XEROX LOST IT in IEEE Spectrum

The inside story of personal computing at the legendary research lab

TEKLA S. PERRYPAUL WALLICH04 JUN 2022N LATE 1969, C. Peter McColough, chairman of Xerox Corp., told the New York Society of Security Analysts that Xerox was determined to develop “the architecture of information” to solve the problems that had been created by the “knowledge explosion.” Legend has it that McColough then turned to Jack E. Goldman, senior vice president of research and development, and said, “All right, go start a lab that will find out what I just meant.”

This article was first published as “Inside the PARC: the ‘information architects’.” It appeared in the October 1985 issue of IEEE Spectrum. A PDF version is available on IEEE Xplore. The diagrams and photographs appeared in the original print version.

Goldman tells it differently. In 1969 Xerox had just bought Scientific Data Systems (SDS), a mainframe computer manufacturer. “When Xerox bought SDS,” he recalled, “I walked promptly into the office of Peter McColough and said, ‘Look, now that we’re in this digital computer business, we better damned well have a research laboratory!’ ”

In any case, the result was the Xerox Palo Alto Research Center (PARC) in California, one of the most unusual corporate research organizations of our time. PARC is one of three research centers within Xerox; the other two are in Webster, N.Y., and Toronto, Ont., Canada. It employs approximately 350 researchers, managers, and support staff (by comparison, Bell Laboratories before the AT&T breakup employed roughly 25,000). PARC, now in its 15th year, originated or nurtured technologies that led to these developments, among others:

The Macintosh computer, with its mouse and overlapping windows.

Colorful weather maps on TV news programs.

Laser printers.

Structured VLSI design, now taught in more than 100 universities.

Networks that link personal computers in offices.

Semiconductor lasers that read and write optical disks.

Structured programming languages like Modula-2 and Ada.

In the mid-1970s, close to half of the top 100 computer scientists in the world were working at PARC, and the laboratory boasted similar strength in other fields, including solid-state physics and optics.

Some researchers say PARC was a product of the 1960s and that decade’s philosophy of power to the people, of improving the quality of life. When the center opened in 1970, it was unlike other major industrial research laboratories; its work wasn’t tied, even loosely, to its corporate parent’s current product lines. And unlike university research laboratories, PARC had one unifying vision: it would develop “the architecture of information.”   ..... ' 

Effectively Organizing Information

A considerable claim to do this effectively. Out of USC ISI.  Looking for more info.

‘Information Is Powerful, But Only If We Can Organize It Right’ — New AI Potential Boon to Businesses

Greg Hardesty | January 19, 2022  in USC Viturbi

Algorithm created by USC’s Information Sciences Institute (ISI) helps e-commerce companies better organize their products and makes it easier for customers to find what they want.

E-commerce sites are in a constant struggle to prevent from drowning in an ever-growing tsunami of information about the products they sell.

How can they best keep a handle on all that information, and, on the consumer side, how can potential customers make sure they can quickly and accurately get what they’re looking for?

The answer is by creating an effective and efficient taxonomy, a fancy word for a system used to classify information in a tree-like structure – think of the classic example of a company’s organizational chart listing the bigwig at the top and the titles of the people below indicating who reports to who.

That’s easy-peasy.

But when it comes to e-commerce, things can get complicated very quickly. New versions of products come out all the time, and people sometimes use different names to describe the same products.

The ongoing challenge for mega-retailers like Walmart and Target, as well as smaller companies, is to create and maintain a taxonomy that keeps their products organized in the most efficient way and makes it easy for website visitors to navigate and find what they want. The information the big players manage is staggering: According to Google, Amazon currently sells more than 350 million products (including third-party sellers) and Walmart currently sells 75 million.

Mayank Kejriwal

MAYANK KEJRIWAL OF THE USC VITERBI SCHOOL OF ENGINEERING

USC Viterbi’s Mayank Kejriwal, a research assistant professor in the Daniel J. Epstein Department of Industrial and Systems Engineering and a research lead at the USC Information Sciences Institute, has created an algorithm he says allows e-commerce and other web-based companies to quickly and cheaply build a taxonomy that can be easily customized to their needs.

Think of the tool this way, he said: Imagine you’re a kid. Now imagine you’re given thousands of pieces of paper with an item written on each piece, such as “baby powder,” “Coca-Cola,” “PlayStation” and “T-shirt.”

Now, suppose you’re asked to build a “tree” out of these pieces of paper so that you could easily find any item when asked.

How long would it take you to build that tree?

“Our system does it in seconds,” Kejriwal said, “and our trees are of similar quality to any that you might be able to build.”

Such lightning-fast speed can save companies money, he added.

The AI tool, Kejriwal explained, has the potential to benefit large advertising companies like Google and media companies that have to “match” product categories to customers so they can get to the right websites, as well as aggregators like eBay and Pinterest where there are many independent third-party sellers. Even if such companies build a taxonomy manually, it would be constantly changing.  .... '

Google Discusses How it Organizes Information

Below the introduction.   Informative.

How Google organizes information to find what you’re looking for

By Nick Fox in the Google Blog

Vice President of Product & Design, Search and Assistant

When you come to Google and do a search, there might be billions of pages that are potential matches for your query, and millions of new pages being produced every minute. In the early days, we updated our search index once per month. Now, like other search engines, Google is constantly indexing new info to make accessible through Search.

But to make all of this information useful, it’s critical that we organize it in a way that helps people quickly find what they’re looking for. With this in mind, here’s a closer look at how we approach organizing information on Google Search.

Organizing information in rich and helpful features

Google indexes all types of information--from text and images in web pages, to real-world information, like whether a local store has a sweater you’re looking for in stock. To make this information useful to you, we organize it on the search results page in a way that makes it easy to scan and digest. When looking for jobs, you often want to see a list of specific roles. Whereas if you’re looking for a restaurant, seeing a map can help you easily find a spot nearby. 

We offer a wide range of features--from video and news carousels, to results with rich imagery, to helpful labels like star reviews--to help you navigate the available information more seamlessly. These features include links to web pages, so you can easily click to a website to find more information. In fact, we’ve grown the average number of outbound links to websites on a search results page from only 10 (“10 blue links”) to now an average of 26 links on a mobile results page. As we’ve added more rich features to Google Search, people are more likely to find what they’re looking for, and websites have more opportunity to appear on the first page of search results.   ..."