Innovation Profs - 11/7/2023

Your weekly guide to generative AI tools and news

Innovation Profs
November 07, 2023

Generative AI News

OpenAI turbocharges GPT-4 and makes it cheaper

On Monday, OpenAI announced GPT-4 Turbo, an improvement to GPT-4 that includes an updated knowledge base (including information up to April 2023, while it was previously January 2022) and a much larger context window (128K tokens, roughly equivalent to 300 pages of text), which allows it process and respond effectively much longer inputs. OpenAI also announced lower prices for developers to use GPT-4 Turbo (three times cheaper than GPT-4) and a program called Copyright Shield for ChatGPT Enterprise users, which will provide legal protection for users facing legal charges for copyright infringement.

OpenAI is letting anyone create their own version of ChatGPT

OpenAI also announced on Monday a platform for users to create custom versions of ChatGPT. Each individual AI agent, referred to as a GPT, can be created code-free using a conversational interface. Creators can monetize their own GPTs on the yet-to-be-released GPT Store. This feature will be available to ChatGPT Plus and ChatGPT Enterprise users.

Brave responds to Bing and ChatGPT with a new ‘anonymous and secure’ AI chatbot

Users of Brave, a privacy-focused browser, can now access Leo, a new AI assistant which offers “unparalleled privacy.” Leo was built using Meta’s open source language model Llama 2. A paid version of Leo uses Claude Instant, a lightweight version of Anthropic’s Claude 2. The LLM ecosystem continues to grow evermore intertwined!

Stability AI brings advanced 3D and image fine-tuning to Stable Diffusion

Stability AI, the company behind the AI image generation platform Stable Diffusion, has announced Stable 3D, a new text-to-3D image tool. The tool will provide “functionality that could help with any type of 3D content creation, including graphic design and even video game development.”

Researchers Say NSFW AI Images Can be Generated by Nonsense Prompts

Researchers at John Hopkins University found that feeding non-sensical prompts to AI image generation tools such as DALL-E 2 and Stable Diffusion resulted in NSFW content. Clearly there’s still more work to be done on the content moderation front for these tools.

Quick Hits

How Chinese influencers use AI digital clones of themselves to pump out content

Questions over honesty and legality as livestreamers, particularly in online shopping, use avatars to boost their earnings

www.theguardian.com/world/2023/nov/06/chinese-influencers-using-ai-digital-clones-of-themselves-to-pump-out-content

GPT-4 falls short of Turing threshold

One question has relentlessly followed ChatGPT in its trajectory to superstar status in the field of artificial intelligence: Has it met the Turing test of generating output indistinguishable from human response?

techxplore.com/news/2023-11-gpt-falls-short-turing-threshold.html

How AI chatbots like ChatGPT or Bard work – visual explainer

In the last year, chatbots powered by large language models are everywhere and even … useful. But how do they work?

www.theguardian.com/technology/ng-interactive/2023/nov/01/how-ai-chatbots-like-chatgpt-or-bard-work-visual-explainer

Ubisoft Using AI-Generated Assassin’s Creed Art Amid Cost Cutting

Knock-off Ezios appear on Ubisoft’s social media channels

kotaku.com/assassin-s-creed-ai-ezio-art-ubisoft-1850977102

A Tool to Supercharge Your Imagination

New technologies for making pictures can be prosthetics for your mind.

www.theatlantic.com/technology/archive/2023/10/ai-image-generation-human-creativity-imagination/675840

Inside the Magical World of AI Prompters on Reddit

The story of AI images isn’t just about robots, but also about the people behind these strange, futuristic visuals.

hyperallergic.com/855052/inside-the-magical-world-of-ai-prompters-on-reddit

This is what happened when Julia Louis-Dreyfus used AI to write an acceptance speech

ChatGPT apparently mistook her for actor Julia Roberts.

www.today.com/popculture/news/julia-louis-dreyfus-chat-gpt-rcna123925

Tool of the week: Grok

Elon Musk announced the release of chatbot Grok, a generative AI tool trained on X / Twitter content. Musk said Grok is being trained by real-time access to information from the platform.

The tool is still in the early stages, but it will soon be available to X premium subscribers.

Innovation Profs Homework

Try out Angry Pumpkins below. Or better yet, make your own version of the game.

AI-generated image of the week

The corn harvest continues here in Iowa, even in our generative AI world…

Prompt: a combine during fall harvest in iowa, the sun sets in the background casting long shadows over the cornfield, 4k, realistic, wide shot, photograph

Generative AI tip of the week

How do we ensure the human touch has a place in AI-generated art? The article linked below has some helpful steps to follow: (1) ideate, (2) iterate, (3) de-pixalate, and (4) manually create. Take a look!

AI-Generated Art Still Needs a Human Touch

Dall-E 2, Midjourney, and Stable Diffusion produce impressive images on command, but perfecting them requires patient, skilled tending.

www.worth.com/ai-generated-art-still-needs-a-human-touch

Get starting with Generative AI

New to generative AI? Here are some places to start…

What we found: Angry Pumpkins

Javi Lopez (@javilopen) created Angry Pumpkins - a version of Angry Birds - using ChatGPT, Midjourney and DALL-E. You can play his version here (doesn’t work on mobile) and see how he made the game here.