Innovation Profs - 12/17/2024

Your weekly guide to generative AI tools and news

Author

Innovation Profs
December 17, 2024

Generative AI News

More of OpenAI’s 12 days of ‘shipmas’!

Last week, we highlighted the first three days of the 12 Days of OpenAI. This week, we have five more days of surprises to report.

ChatGPT with Canvas, a tool introduced back in early October that allows users to make inline edits to ChatGPT’s outputs and to work collaboratively with ChatGPT to modify highlighted sections of the output text (with different sets of tools for word processing and coding), is now available for all users. Another new feature of Canvas is the ability to run the code generated in a given canvas, although we have yet to find an example where the code compiled without throwing an error.

Apple’s partnership with OpenAI, to use ChatGPT as the AI system powering iPhones, iPads, and Mac computers, was previously announced this past summer. The fifth day of OpenAI’s “shipmas” festivities finally brought the concrete fruits of this partnership, with ChatGPT being integrated into Siri and Apple’s new Writing Tools feature. An OpenAI account is not required to access these features.

Yet another raincheck was cashed by OpenAI this past week. When GPT-4o was announced back in May, it was demoed with a real-time video integration, allowing the model to process video inputs to inform its outputs. And yet, that feature has been nowhere to be found…until now. The feature, Advanced Voice Mode with vision, was also rolled out with “Santa mode,” which allows users to converse with an AI simulation of Santa Claus.

ChatGPT users may not be clamoring to try out Santa mode, but they might flock to ChatGPT Projects, the seventh Christmas morsel dropped by OpenAI last week. Projects allows users to store notes, files, and previous conversations, all of which will inform conversations that one has with ChatGPT within a given project, thus removing the need to re-upload the relevant background information in each new session.

Finally, just yesterday OpenAI announced that ChatGPT Search, a feature that has only been available to paid ChatGPT users over roughly the past six weeks, is now available for all users, enabling them to search the web within a ChatGPT query and receive a list of websites as output. We tried searching for “des moines ice skating” and received a snazzy pinned map of the Des Moines metro that included links to six options. The search feature also works with Advanced Voice Mode.

By the end of the week, we’ll send out an update of the last four day of the 12 Days of OpenAI…

Quick Hits

Gemini 2.0 is here and is a true generational leap forward in AI

Remember when Google promised us an AI revolution? Well, hold onto your hats, because Gemini 2.0 just landed, and it's bringing a whole new level of "wow" to the table. This isn't just about asking your AI assistant to set a timer anymore – we're talking about AI that can reason, plan, and even interact...

chromeunboxed.com/gemini-2-0-is-here-and-is-a-true-generational-leap-forward-in-ai

What exactly is an AI agent? | TechCrunch

Regardless of how they're defined, the agents are for helping complete tasks in an automated way with as little human interaction as possible.

techcrunch.com/2024/12/15/what-exactly-is-an-ai-agent

The AI revolution is running out of data. What can researchers do?

AI developers are rapidly picking the Internet clean to train large language models such as those behind ChatGPT. Here’s how they are trying to get around the problem.

www.nature.com/articles/d41586-024-03990-2

'Reasoning' AI models have become a trend, for better or worse | TechCrunch

So-called reasoning models like OpenAI's o1 look to be the next trend in AI. There's pros and cons to this, experts say.

techcrunch.com/2024/12/14/reasoning-ai-models-have-become-a-trend-for-better-or-worse

Pika 2.0 launches in wake of Sora, integrating your own characters, objects, scenes in new AI videos

Advanced image recognition tech ensures these components are seamlessly integrated into AI generated videos, giving creators more control.

venturebeat.com/ai/pika-2-0-launches-in-wake-of-sora-integrating-your-own-characters-objects-scenes-in-new-ai-videos

AI Thinks Differently Than People Do. Here’s Why That Matters.

Generative AI isn’t the strategic oracle many say it is.

hbr.org/2024/12/ai-thinks-differently-than-people-do-heres-why-that-matters

Character.AI has retrained its chatbots to stop chatting up teens

It’s part of a larger suite of changes to the service.

www.theverge.com/2024/12/12/24319050/character-ai-chatbots-teen-model-training-parental-controls

I just saw the future of the web — Google's new Deep Research tool unleashes swarms of AI agents to do in-depth research for you

A Gemini feature that has to be seen to be believed

www.tomsguide.com/ai/google-gemini/i-just-saw-the-future-of-the-web-googles-new-gemini-deep-research-ai-agent-is-incredible

Tool of the week: Sora

We finally got access to OpenAI’s Sora video model, and the results are …. interesting. Watch how these dogs morph into different dogs mid-video. My youngest son wanted a video of his Elf on the Shelf hugging a snow man.

Free users can generate up to 20 five-second videos per month at 720p resolution.

AI-generated image of the week

‘Tis the season to make Christmas cockroaches in Midjourney. Be sure to ask ChatGPT to tell you the story of the Christmas cockroaches. 

Prompt: Christmas cockroaches

Get starting with Generative AI

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

What we found

Whisk is the latest experiment from Google Labs. Users create or upload a subject, scene and style. Then, Whisk brings the three things together, including the ability to remix them to create something unique, from a digital plushie to an enamel pin or sticker. Try it here.