Gen AI News Summary 04.11.24

A bracing sail through the seas of generative AI, handily mapped out under the following headings:

– AI and Content Creation

– AI Models and Tools

– AI and Search

– AI and Avartars

– AI and Agents

– AI Marketing and Sales

– AI and Retail

– AI Health and Education

– AI Adoption

– AI Regulation and Ethics

– And Finally…


AI and Content Creation

Stable Diffusion 3.5
Stability AI’s Stable Diffusion 3.5

Midjourney’s ‘Powerful’ AI image editor now lets you edit any image – including images from the web.

More on Midjourney’s new edit any image feature

How to use Midjourney’s new edit any image feature

Stability AI has updated its image model with several new model variants, including Stable Diffusion 3.5 Large, Stable Diffusion 3.5 Large Turbo, and Stable Diffusion 3.5 Medium. These models are highly customisable, run on consumer hardware, and are free for both commercial and non-commercial use under the permissive Stability AI Community License.

More on Stable Diffusion 3.5

Recraft‘s newest image model – the mystery AI that beat Midjourney and DALL-E in anonymous evaluations  anonymous evaluations – can generate high-quality images with impressive details, quality, and prompt fidelity. And finally, there’s an image generator that also does text well!

More on Recraft’s impressive new image model

Respected AI image start up Ideogram AI launches an ‘infinite canvas’ feature (not unlike the recently introduced GPT-4o with canvas). Ideogram canvas users can spread newly generated images out, compare them to older generations, resize and reorder them at will, and even combine multiple AI generated images into one new composite.

More on Ideogram’s new canvas feature

Since the Adobe Firefly Gen AI updates after last month’s Adobe MAX conference, some users have found that Adobe’s Gen AI tools in Adobe Camera Raw, Lightroom, and Photoshop have become less accurate.

More on Adobe’s AI recent performance

The Apple Intelligence” roll out may have underwhelmed so far, but is Apple’s AI photo “clean up” tool better than Adobe’s? Some think so…

More on Apple Intelligence photo clean up tool

Canva launches Dream Lab, a powerful AI image generator for creatives, developed on top of the gen AI model Leonardo.Ai that Canva acquired earlier this year.

More on Canva

Meta’s Gen AI image model is being lauded for being easy to use. One simple prompt can make AI images “come alive” in Facebook Messenger and Instagram, apparently.

More on AI images in Facebook Messenger and Instagram

ElevenLabs now lets you create your very own custom voiceover voices from text prompts.

More on ElevenLabs new custom voiceovers

Jacob Collier, a Grammy-winning musician, has teamed up with Google DeepMind and Google Labs to create MusicFX DJ, an AI-powered music tool. The interface has been redesigned to encourage creativity and help users easily enter a “flow state” of artistic inspiration. MusicFX DJ is available now, “offering intuitive controls for all skill levels”.

More on MusicFX DJ


AI Models and Tools

OpenAI's ChatGPT 5. Coming soon?
OpenAI’s ChatGPT-5. Coming soon?

Oh yes they will…

OpenAI plans to release its next big AI model by December. A report revealed that OpenAI would release its new ‘Orion’ frontier model (ie GPT-5, or whatever it will be called) by December, with Microsoft and other huge companies getting access before individuals.

More on OpenAI’s plans to release the next version of ChatGPT

And, then another report citing OpenAI CEO Sam Altman, said…

Oh no they won’t!

OpenAI CEO, Sam Altman, responded directly to the report on X, posting “fake news out of control”. An OpenAI spokesperson clarified that they have no plans for an “Orion” release this year but plan to release “a lot of other great technology.”

OpenAI introduces an open source “factuality benchmark” to measure the factual accuracy of language models (the likelihood that they won’t hallucinate). The new benchmark is called SimpleQA.

More on OpenAI’s factuality benchmark

The Perplexity app for the Apple Mac desktop was released, making it more convenient to use the “Google search for research killer”, if you use a Mac…

More on Perplexity’s new Mac desktop app

Not to be outdone, Anthropic also release a desktop app for Claude, for both Mac and Windows.

More on Claude’s new desktop apps

Anthropic has also added PDF support to its Claude 3.5 Sonnet AI model in public beta, allowing it to process both the text and the images within PDF documents.

More on Claude and PDF support

After reports that OpenAI is planning to launch the next version of its flagship AI model in December, there is now a possibility that Google may be planning to launch the latest version of Gemini – Gemini 2.0 – in the same month.

More on the possible December launch of the next version of Gemini

Google is building controls into Gemini, so that Google smart home devices can be controlled with natural languages.

More on Gemini smart home device controls

Apple finally launches Apple Intelligence, with the release of iOS 18.1 if you have a new enough iPhone, iPad or Mac, and you’re happy to set it to US English, for the moment. This initial version introduced a more natural-sounding Siri, major upgrades for Apple’s Photos app, including a new “Clean Up” tool, and systemwide Writing Tools to help users rewrite, proofread, and summarise text in apps like Mail, Messages and Notes. But it currently lacks the conversational abilities we’ve come to expect from an AI assistant, the ChatGPT integration and user-created Genmoji features that many were expecting. Underwhelming, seems to be the general verdict. But, don’t bet against Apple, a latter day tortoise in the race against the hare…

More on the launch of Apple Intelligence

Apple Intelligence now available
First version of Apple Intelligence now available

Meta has struck a multi-year deal with Reuters to use its news content to provide real-time answers to user queries about news and current events in its AI chatbots.

More on Meta’s deal with Reuters

Meta released new versions of its Llama 3.2 AI models that run up to four times faster and achieve a 56% reduction in model size compared to their original counterparts. These breakthroughs make it more feasible to run powerful AI features directly on a mobile phone.

More on Meta’s new Llama 3.2 models

Elon Musk-owned xAI has added image-understanding capabilities to its Grok AI model. This means that paid users on his social platform X, who have access to the AI chatbot, can upload an image and ask the AI questions about it.

https://techcrunch.com/2024/10/28/xai-adds-image-understanding-capabilities-to-grok

AI and Search

OpenAI launch ChatGPT search
OpenAI launch ChatGPT search

OpenAI and search part I: OpenAI launches its web search engine, as a feature in ChatGPT, initially to premium users, and then to enterprise, education and free users in the coming weeks.

More on the roll out of the ChatGPT web search feature

OpenAI’s own summary of the ChatGPT web search feature

OpenAI and search part II: ChatGPT now lets you search your old chats in the web app. OpenAI says “Only you have access to your conversation history, and OpenAI doesn’t use these conversations for training unless you explicitly consent by opting in.”.

More on OpenAI’s new conversation search feature

Meta is reportedly developing its own search engine, to reduce its dependence on Google search and Microsoft Bing.

More on Meta developing its own search engine


AI and Avatars

D-ID's  new real-time conversation avatars
D-ID’s new real-time conversation avatars

HeyGen rolled out new “Interactive Avatars” to allow you to have personalised AI-driven, “immersive, real-time conversations”. Users can either select a template avatar or create their own avatar for a specific use by selecting the option “All Avatars”.

More on creating a HeyGen interactive avatars

Meanwhile…. D-ID launched new high-quality avatars capable of real-time conversations.

More on D-ID “real time conversation” avatars


AI and Agents

KPMG is developing AI agents
KPMG is developing AI agents

Google is developing a computer-using agent – AI that can take over your web browser to complete tasks such as gathering research, purchasing a product or booking a flight. The product, code-named Project Jarvis, is thought to be similar to one Anthropic has just announced. Google plans to preview the product as early as December alongside the release of its next flagship Gemini large language model.

More on Google developing a computer-using agent

Move over Salesforce , Microsoft et al, Big Four accounting and consulting firm KPMG is developing AI agents and is interested in becoming a leader in the emerging AI agent space.

More on KPMG developing AI agents

AI agents will be at centre of our digital worlds, dancing across our devices from smart glasses to cars, providing a consistent experience and adapting the way technology interacts with us.

More on the approaching agentic AI world

OpenAI is expected to launch agents in 2025. Salesforce’s CEO announced AI agents are the third wave of AI. Microsoft adding agent capabilities to Copilot. The message here is clear: AI agents are going to be big, and leaders need to begin strategising how to incorporate this powerful technology into their organisations.

More on why business leaders need to think about AI agents


AI Marketing and Sales

Amazon Ads AI Creative Studio
Amazon Ads AI Creative Studio

A look at how leaders can maximise AI-driven sales strategies.

More on AI driven sales strategies

Harvard Business Review asks: “Can startups thrive in an age of AI?

More on how AI is transforming the start-up landscape

Amazon announces new image, audio and video AI-powered tools for marketers making ads, as part of an AI strategy, that drove up Amazon capital expenditures 81% year on year in Q3.

More on Amazon’s AI-powered advertising tools


AI and Retail

Perplexity is building an AI-powered shopping experience.
Perplexity is building an AI-powered shopping experience.

Perplexity is quietly planning to take on Amazon by building an AI-powered shopping experience. The new ‘Pro Shop’ feature allows users to shop on Perplexity without leaving the platform.

More on Perplexity’s plans for AI-powered shopping


AI Health and Education

NHS England to trial  AI tool to predict patients’ risk of  heart disease
NHS England

NHS England is to trial an AI tool that can predict patients’ risk of developing heart disease, and their risk of early death, using an electrocardiogram (ECG).

More on an AI tool that can predict the risk heart disease

A new deep learning model, developed by the University of Texas Southwestern Medical Center, could lead to more timely and accurate cancer assessments, helping many patients avoid unnecessary surgery and improve outcomes.

More on a new AI model that could reduce the need for surgery for cancer

Biotech startup Iambic Therapeutics just revealed Enchant, an AI platform designed to predict how drug candidates perform in human trials before leaving the lab.

More on AI platform that predicts clinical outcomes from drug discovery

The parents of a high school senior in Massachusetts argued in court that their son was unfairly punished for using artificial intelligence while researching a history project, harming his prospects for acceptance to an elite college.

More on parents sue school for unfair AI accusations

A new research report by Common Sense Media found that about two thirds of the parents of kids who are using AI are oblivious to that fact. And, nearly half said they hadn’t spoken with their teenage kids about AI.

More on kids, parents and AI use research report

The South Korea Ministry of Education plan to integrate AI into the public education system using digital textbooks that leverage AI to personalise learning experiences for each student.

More on the South Koreans approach to using AI with students


AI Adoption

Ethan Mollick, The Wharton School
Professor Ethan Mollick, The Wharton School

The Wharton School professor Ethan Mollick says companies must make organisational changes if they want to benefit from AI.

More on Ethan Mollick and corporate AI implementation

The Generative AI landscape shifted dramatically in 2024, according to a new research study. Nearly three in four executives, 72%, report using gen AI at least once a week, up from 37% in 2023, according to a new study by AI at Wharton, a research centre at the The Wharton School of the University of Pennsylvania, in collaboration with GBK Collective, reveals a dramatic rise in Gen AI adoption across key business functions, as companies move from cautious exploration to rapid integration.

More on “AI at Wharton” study on Gen AI adoption

Microsoft Copilot AI use extends deep into corporate America, but companies are not 100% sold.

More on US corporate adoption of MS Copilot


AI Regulation and Ethics

X.AI will train Grok on your data
X.AI

Elon Musk’s xAI uses all your Twitter/X posts to train its AI model Grok…

More on XAI training Grok on your X/Twitter posts

Google open-sourced its watermarking tool for AI-generated text

More on Google’s open source AI watermarking tool for text

Google announced it will add a note to photos people edit with AI tools, such as Zoom Enhance, Magic Eraser and Magic Editor, to aid transparency.

More on Google’s AI photo edit note

Several researchers raised concerns after finding that OpenAI ‘s Whisper transcription tool suffers from frequent hallucinations and invents text that never appears in recordings despite being deployed extensively in healthcare settings. Over 30,000 medical professionals use Whisper-based tools despite OpenAI’s warnings against high-risk applications, according to a The Associated Press report.

More on the inaccuracy of OpenAI’s Whisper transcription tool

Biden Administration issues first ever national security memorandum on artificial intelligence.

More on Biden’s Memorandum on AI

Chinese research institutions with ties to the Chinese People’s Liberation Army used Meta’s open-source Llama artificial intelligence model to develop an AI tool with potential military applications, Reuters reported, raising further concerns over how China’s government uses open-source AI models from U.S. companies to expand its military and intelligence capabilities.

More on Reuters’ report on China’s army’s use of Meta’s open source models

Big Tech Is paving the way for a nuclear power breakthrough. Small modular reactors, made commercially viable by AI processing needs of AI, could eventually make the power source cheaper, safer and faster to build

More on AI companies use of nuclear power


And finally…

Perplexity's dedicated hub for U.S. general election information.
Perplexity’s dedicated hub for U.S. general election information

Perplexity announced a dedicated hub for U.S. general election information. Populated by data from The Associated Press and Democracy Works, the company described it in a blog as “an entry point for understanding key issues.”

More on Perplexity’s US Election hub


Leave a Reply