 Google announced Bard, a conversational AI service that's likely to compete with ChatGPT. Bard, an experimental technology, is opening to a group of testers today before a wider release to the public in the "coming weeks," wrote Google CEO Sundar Pichai in a blog post. More: - Like ChatGPT, Bard can participate in conversations and answer questions.
- Bard is based on Google's LaMDA, or Language Model for Dialogue Applications, one of the company's large language models.
- Pichai said Bard can draw in internet data. In a demo, Bard answered a query by supplying three bulleted answers about the James Webb Space Telescope.
- According to Pichai, Google is using a LaMDA version that doesn't need as much computing power. This allows it to accommodate more users as the company gains feedback on the system.
Quotable: - Pichai described Bard as "an outlet for creativity, and a launchpad for curiosity, helping you to explain new discoveries from NASA's James Webb Space Telescope to a 9-year-old, or learn more about the best strikers in football right now, and then get drills to build your skills."
THE NEW YORK TIMES 
| |
Live streaming platform Twitch has suspended an AI-generated Seinfeld parody after one of the characters made transphobic remarks. The AI streaming show "Nothing, Forever" was trained on TV episodes to mimic Seinfeld. More: - One of the sitcom's characters, "Larry," who is modeled after Jerry Seinfeld, made transphobic remarks during an AI-scripted standup act.
- Shortly after the segment aired, Twitch removed the show and has now suspended it from the platform for at least 14 days.
- According to the program's creators, the text was generated due to a switch in AI models.
- The show experienced an outage with OpenAI's GPT-3 Davinci model. As a result, the team switched to OpenAI's previous model, Curie.
- "We leverage OpenAI's content moderation tools, which have worked thus far for the Davinci model, but were not successful with Curie," the creators explained.
- The team said it is working to integrate OpenAI's content moderation API before relaunching "Nothing, Forever." It's also looking into backup content moderation systems.
THE GUARDIAN
| |
 A message from TRUERA Free workshop on AI/ML quality and performance Are you a data scientist or machine learning engineer interested in learning more about how to analyze, test, and improve the performance and explainability of your machine learning models? Then this 4-session live online workshop is for you! Register today – space is limited! AI Quality: Driving ML Performance and Trustworthiness is a free class taught live by five experts from leading universities, including Carnegie Mellon University, MIT and Stanford University. Each session covers the essentials of key topics for managing machine learning models, including: - Accuracy and performance debugging
- Model drift
- Fairness
- NLP model quality
- ML explainability
Dates: March 2, 9, 16 and 23 Where: Virtual Event - Join from anywhere Who should attend: Data scientists, machine learning engineers, data analysts, statisticians, data and analytics managers Register for free
| |
Stock photo platform Getty Images has once again sued Stability AI, the creator of the Stable Diffusion image generator, alleging copyright infringement. The complaint — filed on Friday in the U.S. District Court for the District of Delaware — comes after Getty sued Stability in the U.K. last month. More: - In the latest lawsuit, Getty accused Stability of illegally using more than 12 million images and other data to train Stable Diffusion without first obtaining a license or permission.
- The suit alleges that Stability infringed on copyrights and Getty trademarks, noting that its AI system sometimes generates warped versions of the Getty watermark in its outputs.
- Getty has accused Stability of profiting off the photos on its platform and others.
- It has asked the courts to bar Stability from using the images and is also seeking monetary damages.
Zoom out: - A study of 12 million photos from Stable Diffusion's training dataset indicated that more than 15,000 of those photos came from Getty.
- Stable was first trained on a total of 2.3 billion images and then another 600 million for recent checkpoints.
THE VERGE
| |
AI startup Cohere, which specializes in natural language processing (NLP) models for businesses, is in discussions to raise hundreds of millions of dollars, sources told Reuters. While the amount hasn't been disclosed, Reiters report that the funding round would value Cohere at more than $6B. More: - Founded in 2019, Toronto-based Cohere provides an API for companies to deploy natural language processing capabilities, such as chatbots.
- The startup competes with the likes of OpenAI and Anthropic in the generative AI space.
- For example, Cohere plans to soon release a text-generating model that would compete with OpenAI's ChatGPT.
- Cohere's model would be geared more toward businesses and developers, CEO Aidan Gomez told Reuters.
- "Our chat models are focused more on business applicable tasks like answering questions than writing poems," said Gomez, who declined to say how much Cohere is seeking to raise.
Zoom out: - The company has raised at least $170M to date.
- In October, it was reported that Google was in discussions to invest $200M into Cohere.
- Cohere had also communicated with Nvidia about a possible strategic investment.
- Gomez, a co-founder, was previously an intern at Google Brain. He co-authored the paper introducing the Transformer AI model architecture upon which Cohere's technology, OpenAI's GPT-3, and others are built.
REUTERS
| |
 A message from DEEPGRAM Settling for the status quo when it comes to speech recognition? Deepgram is a developer-first speech-to-text platform built with the enterprise in mind. Unlike alternative solutions, Deepgram is built on deep learning to provide faster, more accurate, and more reliable speech recognition. Explore how AI speech-to-text works, common use cases and applications, and how you can get better results with better speech-to-text. NASA, Spotify and Twilio are using Deepgram, download this whitepaper to find out how it works. Get the whitepaper
| |
OpenAI's ChatGPT has the potential to personalize education by catering to each student's learning needs, according to the company's CTO. Mira Murati, who leads the ChatGPT team at OpenAI, told Time magazine that the chatbot "has the potential to really revolutionize the way we learn." More: - In a Q&A with Time, Murati discussed ChatGPT's potential, weaknesses, and the need to govern AI in a way that aligns with human values.
- When asked what problems the chatbot is solving, Murati said it could personalize the learning process. Students could "endlessly converse with a model to understand a concept in a way that is catered to your level of understanding," he said.
- The challenges with ChatGPT, and other similar large luggage models, are that they "may make up facts," Murati noted.
- Given the impact that the technologies will have on society, it's not too early for policymakers to get involved in regulating AI.
- While it's important that OpenAI and other companies release AI technology in responsible and controlled ways, OpenAI is a small organization, and there is more input needed from "regulators and governments and everyone else," Murati said.
TIME
| |
AI Events and Conferences: - World Artificial Intelligence Cannes Festival: Feb. 9-11, Cannes, France, and online.
- Data Science Salon: AI & Machine Learning in the Enterprise: Feb. 21-22, Austin, Texas, and virtual.
- Reinforce AI Conference: March 6-7, Budapest and online.
- ACM/IEEE International Conference on Human-Robot Interaction: March 13-16, Stockholm, Sweden.
- The European Chatbot & Conversational AI Summit: March 15-16, Edinburgh, U.K., and online.
- Memory Con: March 28-29, Mountain View, California.
- Ai4 2023: Aug. 7-9, Las Vegas, Nevada.
- AI Hardware Summit: Sept. 12-14, Santa Clara, California.
- Edge AI Summit: Sept.12-14, Santa Clara, California.
- AI & Big Data Expo Global 2023: Nov. 30-Dec. 1, Olympia, London.
- The AI Summit New York: Dec. 6-7, New York City.
| |
Community Corner I'm Kiani from Inside.com, and last week, we asked our community, "What fields do you see AI not being able to take over?" and received over 100 very split responses! Here are some highlights: - Chris Murphy (p/cemmurphy), VP of Engineering at Nudge Security, wrote, "I think in general AI will touch just about every "field" to some degree. It may be tough, short term, for AI to completely replace humans in fields that require really fine-grained motor skills … It may also be tough for AI to fully participate in fields that require complex, subjective, human interactions." He concluded with, "With that said, I would never ****speak in absolutes. ;-)"
- Patrick Giblin (p/patgiblin), founder of 451 Degrees, wrote, "The better question is what fields do you see AI taking over because there are very few. Influence, yes; take over, no. it is a wonderful tool, not a masterpiece."
- Greg Armstrong (p/greg851602), founder of PencilPay, wrote, "With enough training data and time, it is hard to find something that AI can't do better than a human."
Join the conversations at Inside.com, and we may shout out your name and business or contact you for more! ⭐️
| |
QUICK HITS - Friction is NOT one-size-fits-all. Learn how implementing friction that adapts to your users could boost conversion rates & reduce fraud.*
- Google and Microsoft are on the verge of launching search engines embedded with generative AI, which can supply direct answers to queries rather than only links, according to The Financial Times.
- Recent U.S. trade restrictions have the potential to push back China's AI development by decades, industry insiders say.
- London-based Latent Technology, a startup developing AI technology to improve video game and virtual world animation, secured $2.1M in pre-seed funding led by Root Ventures and Spark Capital.
- Check out this demo-on-demand and find out how 4,000+ fast-growing companies got compliant quickly.*
*This is a sponsored post.
| |
| |
 |  | Beth is a writer and editor at Inside.com covering artificial intelligence and daily technology news. She has written the Inside AI newsletter since 2019. You can reach her at Beth.Duckett@inside.com. |   | | Editor | Aaron Crutchfield is based in the high desert of California. Over the last two decades, he has spent time writing and editing at various local newspapers and defense contractors in California. When he's not working, he can often be found looking at the latest memes with his kids or working on his 1962 and 1972 Fords. | |
Hey data scientists! Join this free workshop on ML model performance & explainability. Register here. | |
| |
Comments
Post a Comment