The Generative AI News (GAIN) rundown for April 6, 2023, focused on regulators and OpenAI, ChatGPT’s popularity compared to the iPhone, deepfake disclosure, authentication and ownership, monetizing those generative AI models, what’s Meta doing, and more.

Bret Kinsella (that’s me) hosts this week with guests Nina Schick, the author of the 2020 book Deepfakes, and Eric Schwartz, head writer at Voicebot.ai. The top stories in generative AI land this week include:

ChatGPT Gets Banned

• A time-out chair for OpenAI and some unfortunate users: Italy took action. Canada opened an investigation. France received complaints. Germany and Ireland indicated they’d like to get involved. Regulators have OpenAI in their sights. How will it go down?

ChatGPT vs. Alexa vs. iPhone

• Compared to what?: ChatGPT is a phenomenon, but how does it stack up to the hype of earlier products? We compare ChatGPT to some notable break-out hits.

Deepfake Solutions

• Provenance in the unreal valley: It’s a deepfake, but you want to disclose its synthetic origins. You also want to show its history and ownership. How about a cryptographic signature from Truepic that tracks the life of the digital artifact?

• The unbearable likeness of your being: Those amazingly lifelike avatars don’t have a clear ownership model today. Someone could make a deepfake of you, and what recourse do you have? However, if you owned the copyright to your digital likeness…

Bing Chat Ads Arrive

• Paying for those GPT-4 inference costs: We knew they were coming, and now we know what they look like, at least one format. Bing Chat has ads that look a lot like what you see in web searches today, with a twist. Generative AI definitely has a revenue model.

Meta Gets Objective

• Alignment is king: Meta rolled out another researcher-only generative AI model. However, this time it showed up with a demonstration app. Segment Anything is a new AI (foundation) model for identifying objects in images and being able to save them separately from the picture with two clicks.

Canva, the True Believer

• Taking the lead over Microsoft: The Redmond giant has talked about DALL-E and GPT-4 in Microsoft Designer and coming to PowerPoint. Canva just started adding new features. A light skepticism from the company in December (ironically about new generative AI features) was replaced by more robust tools and a bigger vision.

More About GAIN

The show is recorded live and streamed via YouTube and LinkedIn at 12 noon ET on Thursdays. You can re-watch each week’s discussion on Voicebot’s YouTube channel. Please join us live next week on YouTube or LinkedIn. Also, participate in the live show by commenting, and we are likely to give you a shoutout and may even show your comment on screen.

The Future of Life Institute, an organization funded by the Musk Foundation, issued a letter calling for a pause of "giant AI experiments" for six months. Elon Musk, Apple co-founder Steve Wozniak, AI legend Yoshua Bengio, and many thousands of others signed the letter.

The idea behind the letter is that the risks posed by AI models such as GPT-4 are potentially so high that we must give policy-makers and technology leaders a chance to assess what guardrails are necessary. But is this a good idea? What are the risks of a pause? What are the objectives and conflicts of interest of the people that signed the letter?

Muddu Sudhakar, the CEO of Aisera, joined me to talk about the letter and all of the discussion it has sparked. We also discussed some alternative approaches, common misunderstandings, and how generative AI is rapidly changing assumptions about our world. 

Sudhakar previously appeared on Voicebot Podcast episode 280. He is a former senior VP and GM at ServiceNow, Splunk, VMWare, and Pivotal. He was CEO at Caspida when the company was acquired by Splunk, where he assumed leadership for machine learning, AI, and analytics-based solutions. Sudhakar was also the CEO and founder of the big data startup Cetas, which was acquired by VMWare, and founded Sanera Systems, which was acquired by Brocade/McData.

He began his career as an engineer at IBM and SGI and earned his PhD in computer science from UCLA. Go Bruins!

The Generative AI News (GAIN) rundown for March 30, 2023, had controversy, competition, Coca-Cola, and more. Is there a more dynamic market right now than generative AI? I don’t think so.

This week’s show is hosted by Bret Kinsella with guests Silke Hahn, technology editor at Heise Online, and Eric Schwartz, head writer at Voicebot.ai. The top stories in generative AI land this week include:

Musk Wants to Slow Down AI

• The Letter: A letter from the Future of Life Institue signed by Elon Musk, Apple co-founder Steve Wozniak, Stability AI founder Emad Mostaque, and 17k others called for a six-month moratorium on “the training of AI systems more powerful than GPT-4.” But maybe this isn’t an entirely altruistic activity.

Google Rumors and More Issues

• Haunted by ChatGPT: Google rumors pointed to the company using ChatGPT data in von-ChatGPT-zum-Training-von-Bard-genutzt-Google-dementiert-8238453.html">Bard training and efforts to combine DeepMind brains and Google AI team brains to catch up with OpenAI’s chatbot. The headlines don’t tell the real story.

OpenAI Declares War and Defends the Citadel

• ChatGPT Plugins: OpenAI is adding plugins to ChatGPT. This puts everyone on notice that OpenAI is now an end-user application product provider and an AI model provider.

• ChatGPT Security: A serious ChatGPT security vulnerability was discovered by a security researcher and promptly patched by OpenAI. There is a bigger story here.

Perplexity AI Lands Funding

• Conversational search wars: It’s not cheap to compete in the search business. Perplexity AI just raised $25.6 million to challenge Bing Chat and Google Bard for conversational search market share.

Stanford Finds Its LLiMit

• Alpaca sent packing: Stanford University launched a web demo of its Alpaca generative AI model based on Meta’s LLaMA model and trained using ChatGPT-generated data. It was shut down in just a couple of days due to safety and cost concerns.

Zoom Goes Generative

• From video to smart companion: Zoom IQ is adding new generative AI features to extend the value of video conferencing beyond the meeting.

Coke Tapping into DALL-E Magic

• New (Coke) custom OpenAI model: Coca-Cola has a new marketing campaign that asks artists to create Coke-themed imagery using a custom implementation of DALL-E.

The show was originally broadcast live on YouTube and LinkedIn, and we also added it to the Voicebot Podcast for your convenience.

Reghu Thanumalayan is a senior vice president at Deutsche Telekom and oversees the Magenta voice assistant. He was my featured guest in Voicebot Podcast Ep 148 three years ago. Reghu joined me to share an update on how things have evolved since launching the product in late 2019.

Magenta has won awards, expanded integrations with TV and smart home devices, and introduced a new call center application. However, the company has also discontinued the smart speaker and learned some tough lessons. 

Reghu breaks down the journey and the learnings. We also talk about a new solution that sounds likely to be a breakout consumer hit and where voice assistants truly excel. 

This interview was conducted onsite at Mobile World Congress 2023 at the Deutsche Telekom booth. 

The Generative AI News (GAIN) rundown for March 23, 2023, was packed with significant announcements. Bret Kinsella hosts this week along with Eric Schwartz. The top stories in generative AI land this week include:

Nvidia Moves up the Stack

• Picasso and NeMo: Nvidia isn’t going just to cash checks for GPU sales related to the generative AI tsunami. They now offer text-to-image and text-to-text models that compete directly with OpenAI, Google, and Stable Diffusion.

• Omniverse Upgrade: Omniverse is an open platform for 3D design collaboration and real-time physically accurate simulations. It announced new services for creating virtual factories mirroring real-world facilities, the option to stream simulated experiences and train to AI-powered robots, and a simulator for testing autonomous vehicles.

Google Teases Gen AI Features and Models

• Show your PaLMs: Google is now letting testers use its giant large language model (LLM) PaLM. This is different from LaMDA and might eventually be a replacement for the brains behind Bard. Right now, it is Google’s answer to GPT-3/GPT-4. A tool to help developers using PaLM called MakerSuite was also announced.

• Gmail and Docs to Get AI writing assistant: Docs and Gmail are getting PaLM-enabled text generation features.

• Bard announced again: Google says it is now offering access to Bard to the general public. But there is a waitlist.

AI-Generated Beauty

• Midjourney 5 is here: There are several minor upgrades, but the key benefits are enhanced quality, more coherence, better photorealism, and more detail.

• Bing adds DALL-E: You can now create images in Bing through a new DALL-E integration. The quality seems better to me than DALL-E 2, which you can access today on OpenAI. Might this be the long-awaited DALL-E 3?

Synthetic Media Policy

• TikTok to tighten deepfake rules: TikTok announced new policies around synthetic media and deepfake use on the platform before its CEO’s Congressional testimony. The policy description grew from 30 words to nearly 400. Will others social platforms take this as a cue to make their own updates?

Games Go Generative

• Roblox wants to make development easier: Roblox added generative AI tools that enable developers to use natural language to create objects and generate code.

• Unity wants generative AI NPCs: Unity didn’t make any concrete announcements. However, its CEO told Reuters that generative AI would help game makers write dialogue and enable non-player characters to interact more naturally with human players.

Other News

• LinkedIn goes generative: The Microsoft-owned company added new generative AI features to create user profiles and job descriptions.

• SoundHound shows Chat AI assistant: The new mobile app offers an assistant that blends SoundHound’s NLU-based assistant with new LLM features.

More About GAIN

The show is recorded live and streamed via YouTube and LinkedIn at 12 noon EST on Thursdays. You can re-watch each week’s discussion on Voicebot’s YouTube channel. Please join us live next week on YouTube or LinkedIn. Also, participate in the live show by commenting.

The Generative AI News (GAIN) rundown for March 16, 2023, required some tough decisions. So much happened this week that we zeroed in on the biggest stories and how they will shape the market. We might get to some of the others, like Midjourney 5, next week. In this episode, Bret Kinsella hosts along with Eric Schwartz. The top stories in generative AI land this week include:

Generative AI for Knowledge Workers

• Microsoft 365 Copilot: A natural language assistant that is a system and not just a bunch of bolt-on LLM features. Can it live up to the ambition? It’s definitely impressive.

• Google Workspace: Some generative AI features are coming to Google Docs and Gmail. Still closed access and little detail, but an official signal about new features. Also, the approach today is less ambitious than Microsoft's.

The Long Wait is Over

• GPT-4 is Here: The much-hyped and long-awaited GPT-4 launch finally arrived. It is multimodal with a vision input element, but that is not widely available. However, the output quality is clearly better, OpenAI claims that factuality is higher and hallucinations rarer, and the context window quadrupled.

ChatGPT Contenders Take a Step Forward

• Anthropic Claude: The ChatGPT competitor was formally announced. However, there is still a waitlist!

• Quora to Monetize Poe: Quora might face an existential crisis as answers become easier to find with ChatGPT, Perplexity, and the New Bing. However, Poe doesn’t rely on one LLM. It allows the user to access many. And Quora is charging for access.

LLMs and Popular Culture

• ChatGPT on South Park: The latest South Park episode featured ChatGPT. When new enterprise technology starts showing up in iconic popular culture venues, something important is happening. Plus, this is just funny.

More About GAIN

The GAIN Rundown was originally broadcast live on YouTube and LinkedIn, and we also added it to the Voicebot Podcast for your convenience. If you would like to watch the show live join us on YouTube or LinkedIn at 12 noon EST on Thursdays. You can re-watch each week’s discussion on Voicebot’s YouTube channel and view the visuals that go along with the show.

The top generative AI news (GAIN) of the week is back for March 9, 2023. This week, Bret Kinsella hosts along with Eric Schwartz from Voicebot.ai and our guest, Brandon Kaplan, chief innovation officer at Journey, and the founder of Skilled Creative. Stories in generative AI land this week include:

Enterprise apps all in on LLMs

• SlackGPT: A new ChatGPT feature for Slack developed by OpenAI

• Salesforce GPT: Einstein GPT features that Salesforce rolled out this week

• Grammarly GPT: Grammarly also adds generative AI features for its 30 million users, putting it on a collision course with Jasper AI, AI21's Wordtune, and Microsoft.

The Evolution of Search

• Answer Box Mania: Both Brave and DuckDuckGo have new search summarization features that appear to replicate the Google answer box. The companies are applying LLMs, but the features are not nearly as ambitious as Bing or Bard. What they do show is how quickly the search experience is changing.

Larger Language Models

• Large, Larger, and Largest: AI21 Labs, a competitor to OpenAI, announced some new, larger, and more polished large language models. The announcement was accompanied by a number of new APIs for a variety of LLM features that developers can access by the drink.

Talk with ChatGPT

• ChatGPT Gets a Face: We also have D-ID's new virtual human-led chat that enables you to have a conversation with ChatGPT-enabled avatar. This discussion included a conversation about the role of virtual humans in interactive bot experiences.

Elon Musk Rumors

• A Singular Generative AI: We finish with a discussion about Elon Musk's plans to challenge OpenAI, a company where he was a co-founder.

The show was originally broadcast live on YouTube and LinkedIn, and you can watch it on Voicebot's YouTube channel.

If you want to get caught up on the top generative AI news of the week, Eric Schwartz and Andrew Herndon from Voicebot.ai and Synthedia break down the top headlines for the first week in March 2023. On tap this week in the video (with links if you want to read more):

• ChatGPT API

• Snap My AI

• Spotify AI DJ

• New Bing in Windows 11

• Meta joins the LLM wars…sort of

More About GAIN

The Generative AI News (GAIN) Rundown is recorded live and streamed via YouTube and LinkedIn at 12 noon EST on Thursdays. Join us live if you can make it. You can re-watch the discussion on Voicebot’s YouTube channel.