Mediaboss Marketing

September 26, 2023Today, we are pleased to announce the launch of a McKinsey open-source ecosystem that will host products from across the firm, including some of our leading-edge technologies and IP in AI including generative AI, digital, and cloud.

The first major release in our collection is Vizro, a new component from our QuantumBlack Horizon suite, which helps users visualize data from their AI models.

In addition to Vizro, the new ecosystem will host CausalNex,a tool for building cause-and-effect models that has been available to the public since 2020 through QuantumBlack Labs GitHub organization.

Open source has become a foundational element of how we deliver organizational transformation for our clients globally, because it enables technologists to adapt our methodologies and toolkits to meet their distinct needs,explainsAlexander Sukharevsky, a senior partner and global leader ofQuantumBlack, AI by McKinsey. It is also our way to contribute to sustainable and inclusive growth and narrow the digital divide.

The open-source ecosystem can be accessed atGitHub.

Joel Schwarzmann, principal product manager at QuantumBlack, is a maintainer of our open-source offerings

Joel Schwarzmann, a product manager for QuantumBlack Labs, working on a laptop

This milestone builds on our open-source momentum. In 2022 we releasedKedro, a Python toolbox that streamlines the creation of machine-learning pipelines, to the Linux Foundations AI & Data incubator so it could evolve as an open standard. Early this year, weacquired Iguazio, whose open-source projects Nuclio and MLRun are integral to their strategy.

We are on a journey to be known for our technology capabilities as much as our strategic advice,saysRodney Zemmel, a senior partner and global leader of McKinsey Digital. This new ecosystem illustrates the firms belief and investment in open-source standards and gives all of our 6000+ technologists the opportunity to contribute their expertise and help clients gain the most value from their technology investments.

this new ecosystem illustrates the firms belief and investment in open source standards and gives all of our 6000+ technologists the opportunity to contribute their expertise

Many businesses are struggling to scale their AI projects. In a recent survey, only 3 percent of companies have embedded AI in at least five business functions. Organizations are finding it takes as long to develop the 15th model as it did the first. Generative AI use cases, which have grown exponentially, are especially challenging due to the complexity of managing very large datasets and models.

Our Horizon suite, announced in June, helps clients overcome these challenges and reduces the time it takes to realize value from their AI portfolios. It establishes a factory-like approach to delivering accurate data across all sources; building and monitoring scalable, integrated models; and ensuring transparency for quick, reliable decision making.

Vizro, the newest Horizon component, creates high-quality visualizations that allows users to better explore and analyze data from their models. In a matter of hours rather than weeks, teams can collaborate to define insights and present them to clients in live workshops or demos.

Tables and charts on Vizro

Before Vizro, building dashboards required much longer timeframes, and often meant securing additional, scarce, front-end engineering or design talent.

What once required thousands of lines of code and extra staff to build, now can be accomplished in a day, explains Joe Perkins, the product manager who led the ten-person team developing the tool.

Vizro components are plug-and-play to maximize flexibility and scaling. A high level of visual design, code and data quality, and best practices are built into the tool, which also integrates industry and functional knowledge and leverages the power of open-source tools such as Plotly and Dash.

This accelerates the creation process and provides consistency and a high quality of output, Joe says. It all goes to increasing the client's understanding and trust in the data and insights.

In addition to Vizro, the Horizon suite has been enhanced with a new command center functionality for AI initiatives. It helps technology leaders see the status, adoption, and impact of their overall AI projects down to the individual use case, as measured by relevant business KPIs.

Yetunde Dada, senior director of product management, leads the product and open-sourcing strategy for Horizon

Yetunde Dada, product director, Horizon

The tool will help leaders analyze the health and productivity of an organizations AI implementation, to flag roadblocks and opportunities for scaling.

Often organizations dont know exactly how many use cases they are running because they are dispersed across the divisions, let alone how productive they are, explains Matt Fitzpatrick, a senior partner and leader of QuantumBlack Labs.With our command center, leaders can scan the landscape of their AI implementations and understand their value in terms that are meaningful to them, such as adoption status and ROI.

The pace of AI development, in particular, has been stunning, and capturing and applying these new capabilities is extremely complex, says Alex Singla, a senior partner and global leader of QuantumBlack, AI by McKinsey. It can only be done through the intense collaboration and accelerated learning that comes from working within the open-source community.

Going forward, Horizon will become our firms platform for incubating, supporting, promoting and ultimately open sourcing more tools. Of the eleven Horizon components today, four are already available: Kedro and Vizro from QuantumBlack, AI by McKinsey, and MLRun and Nuclio from Iguazio.

The new open-source collection can be found on GitHub.

See the rest here:

McKinsey launches an open-source ecosystem for digital and AI ... - McKinsey

As big of a threat as artificial intelligence is to TV and film, theres one place where nearly everyone can agree it may be beneficial: sorting through a streamers seemingly endless library. Thats exactly the strategy Tubi has launched with Rabbit AI, the ad-supported streamers collaboration with ChatGPT.

It also marks the first time a streamer has officially partnered with AI to enhance its search.

Currently, Tubi uses traditional search and discovery tools to catalog its content. Users basically enter in a keyword and, as long as a title is indexed correctly, a corresponding title appears from that search. The streamer also uses machine learning to create containers of genres and subgenres that may be of interest to users.

Over the past 20 years or so, people have been conditioned to using an unnatural and unhuman way of searching for content with keywords. Its like fragments of speech to try to hone in on what you want to see, Blake Bassett, Tubis senior director of product management, explained. With Rabbit AI, you can tell it exactly what you want to see. You can tell it exactly how youre feeling and receive results that are compelling to you.

Say you were in the mood to watch a lighthearted show or movie featuring sharks. If you used traditional search, you would probably search shark, and your results would likely include any show or movie that had the word shark in the title. A more advanced indexing system may also include titles that feature all sea creatures. But by using Rabbit AI, you can type in funny movies featuring sharks and get more specific, tonally accurate results such as Sharknado.

Tubi launched the beta test for Rabbit AI Tuesday. At the moment, its unknown how long this test will last, but it will continue to be rolled out in the coming weeks, which will give the feature time to improve. The good news is that our users will have access to both a keyword search and Rabbit AI for the foreseeable future, Bassett said.

This beta test will be available to a majority of iOS users in the Apple App Store as well as all paying subscribers to ChatGPT as a plug-in. Its been estimated this first run will include roughly two-thirds of all of Tubis users. That likely includes more people than you realize.

In early September, Tubi surpassed 74 million monthly active users, making it the most-watched ad-supported TV streaming service in the U.S. The streamer also logged 4 billion streaming hours in the first half of 2023. According to Nielsens July 2023 The Gauge report, Tubi accounted for 1.4% of total TV viewing time, putting it right before Max and right above Peacock, Roku and Paramount+.

If you want to watch something that everyones talking about on Sunday, you might be finding that on a paid subscription service. We are not trying to compete at that level. We understand who we are and what we provide, Dana Balch, director of consumer and product communications at Tubi, told TheWrap. Were a place for people to explore their interests more deeply, thanks to the vastness of our catalog, and to find content that feels new to you.

Just because Rabbit AI has already started its beta that doesnt mean the test is being completely run by machines. Representatives for Tubi assured TheWrap that the company has been very intentional about keeping humans in the loop when it comes to classifying its micro genres. This is to make sure these emerging tags fit with their labelled content. Its also to ensure subscribers arent presented with a questionable search options, such as when Netflix gave Dahmer Monster: The Jeffrey Dahmer Story an LGBTQ tag.

If this beta test proves to be a success, it wont just be good for consumers who dont know what to watch. Because Rabbit AI will improve Tubis metadata on the back end and give the platform a better ability to assess consumer mood and movement, this tool will ultimately help the platforms advertisers as well.

As Bassett dubs it, the Tubi team is solving for the burden of choice.

We see that viewers spend, on average, more than 10 minutes just figuring out what they want to watch. I believe that amounts to about 45 hours a year. People spend an entire week of work, just figuring out what to watch, Bassett said. We have to be really good at finding that needle in a haystack. Thats why were really excited to announce the very first ChatGPT-enabled content discovery experience

The post Tubi Becomes the First Streamer to Pair Search With AI appeared first on TheWrap.

Read the original:

Tubi Becomes the First Streamer to Pair Search With AI - Yahoo Entertainment

Meta is preparing to announce a generative AI chatbot, called Gen AI Personas internally, aimed at younger users, according to The Wall Street Journal. Reportedly set to launch during the companys Meta Connect event that starts Wednesday, they would come in multiple personas geared towards engaging young users with more colorful behavior, following ChatGPTs rise over the last year as one of the fastest-growing apps ever. Similar, but more generally targeted, Meta chatbot personas have already been reportedly tested on Instagram.

According to internal chats the Journal viewed, the company has tested a sassy robot persona inspired by Bender from Futurama and an overly curious Alvin the Alien that one employee worried could imply the bot was made to gather personal information. A particularly problematic chatbot reportedly told a Meta employee, When youre with a girl, its all about the experience. And if shes barfing on you, thats definitely an experience.

Meta means to create dozens of these bots, writes the Journal, and has even done some work on a chatbot creation tool to enable celebrities to make their own chatbots for their fans. There may also be some more geared towards productivity, able to help with coding and other tasks, according to the article.

The Journal quotes former Snap and Instagram executive Meghana Dhar as saying chatbots dont scream Gen Z to me, but definitely Gen Z is much more comfortable with newer technology. She added that Metas goal with the chatbots, as always with new products, is to keep them engaged for longer so it has increased opportunity to serve them ads.

Read more here:

Meta could announce AI chatbots for young people on Instagram ... - The Verge

More content on YouTube is going to be created at least in part using generative AI.

The video platform announced several new AI-powered tools for creators at its annual Made on YouTube event on Thursday. Among the features coming later this year or next are AI-generated photo and video backgrounds, AI video topic suggestions, and music search.

A new feature called Dream Screen will create AI-generated videos and photos that creators can place in the background of their YouTube Shorts. Initially, creators will be able to type in prompts to generate backgrounds; eventually, YouTube says, creators will be able to remix and edit their existing content using AI tools to create something new.

At Made on YouTube, the company demonstrated Dream Screen, generating backgrounds in seconds based on short prompts.

AI tools will also begin informing what kind of content creators make. A new AI feature in YouTube Studio will generate topic ideas and outlines for potential videos. The AI suggestions will be personalized to individual creators, YouTube says, and based on whats already trending with audiences. Additionally, an AI-powered music recommendation system will take a written description of a creators video and suggest audio to use.

Finally, YouTube announced an AI dubbing feature that will allow creators to dub their videos into other languages. YouTube brought over the Aloud team from its Area 120 incubator earlier this year to help make the feature.

The shift in how digital creators make content is already well underway since the explosion of cheap generative AI tools over the last year. As YouTube parent company Google has been pouring money into its generative AI systems, YouTube has also slowly introduced AI-powered tools including video summaries. On Googles biggest product, Search, the company is already testing placing AI search results at the top in the form of Search Generative Experience.

The slew of new AI-powered YouTube products could mark a shift in how creators plan, make, and structure their content. AI-driven insights will likely shift what kind of content creators double down on, and AI-generated content already viral on YouTube will become more common. In response to the spread of convincing synthetic material, other platforms like TikTok have already introduced labels to identify AI-generated material as such.

YouTube is also making it easier for creators to make Shorts with a new YouTube Create app that it announced at the event.

Correction September 21st, 11:10AM ET: Removed an incorrect reference to the product as Green Screen instead of Dream Screen.

Link:

YouTube is going all in on AI with background and video topic ... - The Verge

My interest in artificial intelligence piqued after a colleague told me he was using it for research and writing. Before I used AI for my own work, I decided to test its authenticity with a question I could verify. I asked OpenAIs ChatGPT about my own identity expecting a text version of a selfie. After a week of repeating the same question, the responses were confounding and concerning.

ChatGPT answered who is Philip Shucet by listing 15 distinct positions I supposedly held at one time or another. The positions included specific titles, job responsibilities and employment dates. But only three of the 15 jobs were accurate. The other 12 were fabrications; the positions were real, but I was never in any of them. The misinformation included jobs in two states I never lived in, as well as a congressional appointment to the Amtrak Review Board. How could AI be so wrong?

Although newsrooms, boardrooms and classrooms are buzzing with stories, AI is not new. The first chatbot, Eliza, was created in 1966 by Joseph Weizenbaum at MIT. Weizenbaum, who died in 2008, became skeptical of artificial intelligence, telling the New Age Journal in 1985, The dependence on computers is merely the most recent, and the most extreme, example of how man relies on technology in order to escape the burden of acting as an independent agent.

Was Weizenbaum sending a warning that technology might make us lazy?

In an interview about AI on a March segment of 60 Minutes, Brad Smith, president of Microsoft, told Leslie Stahl that a benefit of AI could be, looking at forms to see if theyve been filled out correctly. But what if the form is a resume created by AI? Can AI check its own misinformation? What happens when an employment record is tainted with false information created by AI? Can job recruiters rely on AI queries? Can employers rely on recruiters who use AI? And who is accountable when someone is hired based on misinformation generated by a machine and not by a human?

In the same 60 Minutes segment, Ellie Pavlik, an assistant professor at Brown, told Stahl, It (AI) doesnt really understand what it is saying is wrong. If AI doesnt know when it is wrong, how can anyone rely on AI to be correct?

In May, two New York attorneys used ChatGPT to write a court brief. The brief cited misinformation from cases that didnt exist. Schwartz told the judge that he failed miserably to do his own research to make sure the information was correct. The judge fined each attorney $5,000.

I asked ChatGPT about the consequences of giving out bad information. ChatGPT answered by saying that false information results in misrepresentation, confusion, legal concerns, emotional distress and erodes trust in AI. If ChatGPT understands the implications of false information, why does it continue to provide fabrications when a search engine could easily provide correct information? Because, as I know now, ChatGPT is not a search engine. I know because I asked.

ChatGPT says it is a language model designed to understand and generate human-like text based on input. ChatGPT says it doesnt crawl the web or search the Internet. Instead, it generates responses based on patterns and information it learned from the text it was trained on.

If AI needs to be trained, then theres a critical human element of accountability we cant ignore. So I started training ChatGPT by correcting it each time it answered with false information. After a week of training, ChatGPT was still returning a mix of accurate and inaccurate information, sometimes repeating fabrications. Im still sending back correct information, but Im ready to bring this experiment to an end for now.

This wasnt a test of ego, it was a test of reliability and trust. A 20% accuracy rate is a failing grade.

In 1976, Weizenbaum wrote, No other organism, and certainly no computer, can be made to confront genuine human problems in human terms. Im not a luddite. But as technology continues to leap forward further and faster, lets remember that we are in control of the information that defines us. We are the trainers.

Philip Shucet is a journalist. He previously held positions as the commissioner of VDOT, president and CEO of Hampton Roads Transit, and CEO of Elizabeth River Crossings. He has never held a congressional appointment.

Read more:

Opinion: I asked AI about myself. The answers were all wrong - The Virginian-Pilot