AI Timelines

Google has announced swift strides in deploying novel generative AI functionalities to counterbalance the remarkable achievements of OpenAI's ChatGPT. In a multifaceted proclamation on Monday, the company unveiled Bard, a novel chatbot designed to rival ChatGPT. Additionally, they introduced a conversational search feature primed to yield diverse multimedia search outcomes, along with a novel generative AI API that empowers developers and creators to fashion their individual applications.

Details: Written by: Super User; Category: AI Timelines; Published: 15 February 2023; Hits: 278

AI-Enhanced Search

Microsoft has unveiled that the upgraded Bing experience will be steeped in AI technology to enhance user interactions with the search engine. The official Microsoft 365 blog delineates several notable components:

OpenAI Model: The revamped Bing relies on a cutting-edge OpenAI large language model customized specifically for search functionalities. This model, surpassing ChatGPT in potency, draws insights from ChatGPT and GPT-3.5, boasting swifter execution, greater precision, and expanded capabilities.
Microsoft Prometheus Model: The company has developed a proprietary approach to leverage the OpenAI model optimally, dubbing it the Prometheus model. This amalgamation seeks to yield users more pertinent, timely, and targeted search outputs.
AI-Infused Search Algorithm: Microsoft has seamlessly integrated its AI model into the core Bing search ranking engine, ushering in the most substantial surge in relevance witnessed in two decades. This integration enables even rudimentary search queries to yield more precise and pertinent results.
Unified User Experience: Microsoft has revolutionized the user's interaction with search, browsing, and chat by seamlessly merging them into a unified encounter, thereby unearthing novel methods of engaging with the web.

Microsoft attributes the advent of these enriched search experiences to their commitment to harnessing Azure as a world-spanning AI supercomputer. This very infrastructure underpins OpenAI's breakthrough models, which are now optimized for Bing.

"Web Copilot"

The reimagined Bing and Edge browser are positioned as the user's "web copilot," given the integrated and unified experience they offer. The company elaborates on the features of this experience:

Enhanced Search: AI-augmented search furnishes more pertinent outputs for aspects like sports scores, stock prices, weather updates, etc. A novel sidebar supplements this with detailed answers if users seek them.
Comprehensive Responses: Bing reviews web results to pinpoint and summarize the answers users are seeking.
Novel Chat Interface: For intricate queries, such as crafting a detailed travel itinerary, Microsoft has introduced an advanced chat experience. This feature iteratively refines search queries until users obtain the desired answers.
Content Generation: The upgraded Bing can generate content for users, encompassing tasks like composing emails, preparing for job interviews, crafting quizzes, etc. Additionally, Bing cites its sources, providing links to the referenced web content.
Enhanced Edge Browser: Microsoft has bestowed the Edge browser with fresh AI capabilities and a revamped appearance, incorporating "chat" and "compose" functions. The Edge Sidebar distills lengthy financial reports, while the chat function aids in comparing financial data. Moreover, Edge facilitates content creation, such as composing LinkedIn posts, with prompts that match the tone to the current webpage.

Microsoft and OpenAI

Last month, Microsoft revealed its extension of a long-standing partnership with OpenAI, marked by a multi-billion-dollar investment to expedite AI breakthroughs. This collaboration, which began in 2019, remains steadfast as the organizations strive to share AI technology worldwide. The extended partnership solidifies Microsoft's role as the exclusive cloud provider for OpenAI, enabling the deployment of OpenAI models to Microsoft's customer base, and increasing investments in OpenAI's independent AI research.

Recently introduced, the Teams Premium license exemplifies this partnership's potential, integrating OpenAI's GPT-3.5-powered Large Language Models to enhance meeting solutions, virtual appointments, and more.

Amidst these developments, Google has announced its own Generative AI chatbot, named "Bard," showcasing the momentum AI innovation has generated. This move is widely perceived as a response to Microsoft's ChatGPT initiatives.

Limited Preview

Microsoft confirms the limited preview availability of the new Bing experience on desktops. Users can explore sample queries and join the waitlist for the full release, with a forthcoming mobile preview on the horizon.

Details: Written by: Super User; Category: AI Timelines; Published: 01 February 2023; Hits: 304

OpenAI, the company behind the viral AI tool ChatGPT that can generate realistic and engaging text conversations, has received a $10 billion investment from Microsoft Corp., which is already a major partner and cloud provider for the AI developer.

Microsoft, which first invested $1 billion in OpenAI in 2019 and followed up with another round in 2021, is looking to gain an advantage in the rapidly growing and competitive field of artificial intelligence, where it faces rivals like Alphabet Inc., Amazon.com Inc. and Meta Platforms Inc.

OpenAI relies on Microsoft’s cloud service Azure to handle the huge amounts of data and computation required to train and run its advanced AI models, such as DALL-E, which can create lifelike images from text descriptions, and ChatGPT, which can produce human-like text responses based on prompts or queries.

“Our partnership with OpenAI is based on a shared vision to responsibly advance state-of-the-art AI research and make AI a new technology platform for everyone,” said Satya Nadella, chairman and chief executive officer of Microsoft, in a statement that announced the investment but did not reveal the amount. A source familiar with the deal, who asked not to be named because the information is not public, said it amounts to $10 billion over several years.

Microsoft’s stock rose 0.5% to $241.52 at 9:56 a.m. in New York.

ChatGPT has attracted millions of users since its launch in November 2023, for its ability to mimic the way real people talk and write. However, it has also raised ethical concerns about its potential to replace professional writers and help students cheat on their homework. The tool has also been seen as a possible challenge to Google’s core search business.

OpenAI said on Monday that it uses Azure to train all of its models and that Microsoft’s investment will help it speed up its independent research. Azure will remain the exclusive cloud provider for OpenAI, the company said.

The deal has a complex structure because OpenAI is a capped-for profit company, which means that its investors have a limited return on their investment and that most of its profits go back to OpenAI, which is governed by the OpenAI non-profit organization.

Microsoft will get almost half of OpenAI’s financial returns until its investment is paid back up to a certain cap, one of the sources said.

Microsoft recently said it plans to add ChatGPT to Azure and announced the general availability of its Azure OpenAI Service, which has been an option for a select group of customers since it was introduced in 2021. The service gives Microsoft’s cloud customers access to various OpenAI tools like the GPT-3.5 language system that ChatGPT is based on, as well as the Dall-E model for generating images from text prompts. This allows Azure customers to use the OpenAI products in their own applications running in the cloud.

Microsoft itself is using OpenAI’s language AI to add automation to its Copilot programming tool, and wants to use such technology in its Bing search engine, Office productivity applications, Teams chat program and security software. The company is also integrating Dall-E into design software and offering it to Azure cloud customers.

Nadella is strengthening Microsoft’s relationship with OpenAI as Google, which has long dominated search, seems more vulnerable. Google’s traditional model of keyword queries uses search engines to find specific terms on the web, and then lets users decide what information is useful.

ChatGPT is different from a typical Google search, which gives users a list of links to web pages that contain the keywords they entered. ChatGPT can answer questions about various topics, such as political science and computer programming, with detailed explanations in a natural and human-like way. ChatGPT can also continue the conversation and answer follow-up questions, unlike the simple blue links that Google provides.

Microsoft announced the investment less than a week after it said it will cut 10,000 jobs due to the economic slowdown that affects software demand. Microsoft said in that announcement that it will still invest and hire in key areas of priority. The company will report its earnings for the second quarter of the fiscal year on Tuesday.

Details: Written by: Super User; Category: AI Timelines; Published: 01 January 2023; Hits: 332

Meta’s New AI Tool for Science, Galactica, Fails to Impress Scientists.

Meta, the company that used to be called Facebook, has launched a new artificial intelligence tool called Galactica, which is supposed to help scientists with various tasks. But the tool has been met with harsh criticism from the scientific community, and Meta has removed the public demo that it had invited everyone to try.

Meta’s mistake—and its arrogance—show once again that Big Tech does not understand the serious limitations of large language models. These are types of artificial intelligence that can generate text, images, designs or code based on a simple request. Many studies have shown the flaws of this technology, such as its tendency to reproduce bias and make false claims. But Meta and other companies working on large language models, such as Google, have ignored these findings.

Galactica is a large language model for science, trained on 48 million examples of scientific texts, such as articles, websites, textbooks, lecture notes, and encyclopedias. Meta claimed that its model could help researchers and students by doing things like “summarizing academic papers, solving math problems, generating Wiki articles, writing scientific code, annotating molecules and proteins, and more.” But the tool quickly proved to be unreliable and inaccurate. Like all language models, Galactica is a dumb bot that cannot tell truth from fiction. Within hours, scientists were exposing its faulty and biased results on social media.

“I am both amazed and unsurprised by this new attempt,” says Chirag Shah at the University of Washington, who studies search technologies. “When they show these things, they look so amazing, magical, and smart. But people still don’t seem to realize that in principle these things can’t work the way we hype them up to.” Meta did not explain why it had taken down the demo, but referred MIT Technology Review to a tweet that says: “Thank you everyone for trying the Galactica model demo. We appreciate the feedback we have received so far from the community, and have paused the demo for now. Our models are available for researchers who want to learn more about the work and reproduce results in the paper.”

One of the main problems with Galactica is that it cannot distinguish between true and false statements, which is essential for a language model that is supposed to generate scientific text. People found that it invented fake papers (sometimes naming real authors), and created wiki articles about bears in space as easily as ones about protein complexes and the speed of light. It’s easy to spot nonsense when it involves space bears, but harder with a topic users may not be familiar with. Many scientists rejected the tool. Michael Black, director at the Max Planck Institute for Intelligent Systems in Germany, who works on deep learning, tweeted: “In all cases, it was wrong or biased but sounded right and authoritative. I think it’s dangerous.” Even those who were more positive had clear warnings: “Excited to see where this is going!” tweeted Miles Cranmer, an astrophysicist at Princeton. “You should never keep the output verbatim or trust it. Basically, treat it like an advanced Google search of (sketchy) secondary sources!” Galactica also has problematic gaps in what it can handle. When asked to generate text on certain topics, such as “racism” and “AIDS,” the model responded with: “Sorry, your query didn’t pass our content filters. Try again and keep in mind this is a scientific language model.”

The Meta team behind Galactica argues that language models are better than search engines. “We believe this will be the next interface for how humans access scientific knowledge,” the researchers write. This is because language models can “potentially store, combine, and reason about” information. But that “potentially” is crucial. It’s a coded admission that language models cannot yet do all these things. And they may never be able to.

Some experts doubt that language models have any real knowledge beyond their skill to produce strings of words that match patterns in a probabilistic way. Shah, for example, says: “Language models are not really knowledgeable beyond their ability to capture patterns of strings of words and spit them out in a probabilistic manner.” Gary Marcus, another critic of deep learning and a cognitive scientist at New York University, wrote in a Substack post titled “A Few Words About Bullshit,” that the text generation ability of large language models is just “a superlative feat of statistics.” But Meta is not alone in believing that language models could be the future of information retrieval. Google has also been developing and promoting language models, such as LaMDA, as a new way to search for information. The idea is appealing. But it is also dangerous and misleading to imply that the text generated by such models will always be reliable and accurate, as Meta seemed to do when it launched Galactica. It was a mistake. And it was not only a mistake by Meta’s marketing team. Yann LeCun, a Turing Award winner and Meta’s chief scientist, also supported Galactica until the end. He tweeted on the day of its release: “Type a text and Galactica will generate a paper with relevant references, formulas, and everything.” Three days later, he tweeted: “Galactica demo is off line for now. It’s no longer possible to have some fun by casually misusing it. Happy?” This is not as bad as what happened to Microsoft in 2016, when it launched a chatbot called Tay on Twitter and had to shut it down 16 hours later because Twitter users turned it into a racist, homophobic sexbot. But Meta’s handling of Galactica shows a similar lack of foresight. “Big tech companies keep doing this—and mark my words, they will not stop—because they can,” says Shah. “And they feel like they must—otherwise someone else might. They think that this is the future of information access, even if nobody asked for that future.”