Thought #6 - AI Is Changing Fast - So Let’s Get Comfortable Trying Things

Hi lovely humans,

You might know that we’re quite into thinking about learning - and how people can feel confident with AI and data tools. At Taught by Humans, we believe the most important skill for AI is not prompting - it’s experimentation (I’ve written a blog on this if you wanted to read more).

So if you read about new releases or interesting use cases, just start testing things out (with considerations for data privacy and security of course!)

AI New Releases

OpenAI and Google continue to release new models and features every week (but slightly less intense than the previous few weeks):

OpenAI’s new image generation feature

There’s been a lot of buzz about this, and we’re actually quite impressed. You can read about my first attempt using it in our blog.

You can try it for yourself by asking ChatGPT to make an image.

The Ongoing Copyright Discussion - This ability has resulted in a lot of people creating images in the style of the popular Japanese animation company Studio Ghibli - re-sparking the ongoing debate about copyright. Sam Altman (the founder of OpenAI) has even created a Ghibli-style icon of himself and made it his Twitter profile pic (he also commented on the types of images people are generating of him). This contrasts with a now famous 8 year old video of Studio Ghibli co-founder, Hayao Miyazaki, saying AI is “an insult to life itself” (Screenrant).

Just a little reminder that we are still trying to work out where these new shiny tools fit with creativity, ownership, and human values.

Google Gemini 2.5

Google is calling this new model a “thinking model” (this is similar to OpenAI’s o3 reasoning model). As always Google have released this as an experiment to see how it goes.

According to Google, this model is scoring higher on a lot of the standard AI model tests (including Humanity’s Last Exam), but not all of them.

We’ve had a go using the Google AI Studio (it’s not yet available in Gemma, but the AI Studio is easy enough to use if you want to try). Last week we tried Gemini 2 in Gemma, which hallucinated some facts and links about Taught by Humans. Gemini 2.5 on the other hand actually seemed to know who Taught by Humans were and was able to give some useful advice.

Databricks & Anthropic Partnership

Anthropic (the company behind the popular, and my favourite, chatbot Claude) has announced an important partnership with Databricks (an enormous data company - they are known for being a pioneer of Data Lakehouses (a mix between a data lake and a data warehouse) and allow companies to run enterprise level data activities in the cloud).

This means Databricks users can now run Claude 3.7 Sonnet directly within their platform. They say this is democratising AI access by allowing the 10,000+ companies using Databricks to use these top models without the need to code.

AI News

The news this week is mostly about big companies rolling out AI:

• Garmin rolling out AI for top tier customers - The Verge

• H&M using AI models (digital twins of human models) - Business of Fashion

Also worth a read

Not exactly news but we thought these were useful and worth reading:

• A discussion on AI usage in job applications (really relevant to one of our previous polls) - BBC

• Attitudes to AI in UK - an interesting survey detailing the current concerns and thoughts about AI in the UK - Ada Lovelace Institute

LinkedIn AI Poll

Last week we asked how often people are actively using AI - like chatting to a chatbot, using a notetaker, creating images, rather than passively using recommendations on Netflix or getting summaries of meetings without trying.

We had 150 people vote - and (unsurprisingly as it is LinkedIn) over 60% are using AI everyday.

Surprisingly (to me about people on LinkedIn) - 4% are still never using AI!

The comments sparked chat about what people are using AI for - is it for research? Is AI even trustworthy enough for that? So we’ll be exploring that in the next few week’s poll.

Vote in this week’s poll - please!

It’s about how often people are using AI for research

Final Thoughts

As always we hope this was helpful!

Feel free to share this with anyone who might find it useful.

Next week, we’ll start exploring how to support AI experimentation in your teams and businesses.

Laura
Always learning