STRAWBERRY IS HERE
The much vaunted new OpenAI model codenamed Strawberry (a reference to ChatGPTs letter counting issues), with advanced reasoning capabilities, has been released today. It is officially titled "o1-preview". It's unlike all the other GPTs we have become accustomed to because it is much better at thinking through a problem which means it can handle more complex tasks, but also that it takes a lot longer to respond (minutes in some cases). It does mean it is a lot smarter (in human measures) including exceeding "human PhD-level accuracy on a benchmark of physics, biology, and chemistry problems". This doesn't mean it is AGI, or even close to AGI, but it is a big step forward. You can get access to the o1-preview on paid OpenAI subscriptions, but there is no API access as yet.
SO WHAT: This is big news, but it isn't going to have big immediate impacts. It's big because it heralds a new approach to GPTs that makes them more capable at handling complex challenges in a single hit. It seems like it's level of intellect (for lack of a better work) is similar to GPT-4o, but it ends up being smarter because it can reason with itself to drive itself to think through a problem more comprehensively instead of just giving a zero-shot style answer. The effect is that there is less of a need for humans to coach the AI in how to address a problem and that will make it easier to use and deliver better outputs. I can see this being a massive multiplier on multi AI agent setups, albeit we will need to wait for API access to test this out. No word on costs as yet, but I would imagine given the response times, the compute times are much higher and that will translate to a much higher cost than
standard ChatGPT models. So in summary, it's exciting and a big step forward, but its also early days and there are a lot of unknowns and no doubt over the coming weeks it will become clearer just how big of a deal this is.
> Read More
YOU SHOULD READ THIS - GOOGLE'S NOTEBOOKLM AI PODCAST
Google has widely released the previously previewed "Audio Overview" feature to its AI note-taking and research app, NotebookLM. In effect this lets you automatically generate a podcast based on any document or set of knowledge you want - you can even upload multiple documents and it will fold them all into a single podcast. What's more you can also ask the AI any follow up questions on the topic you like and it will give you a customised answer.
SO WHAT: This is amazing. It is a hugely powerful tool, people love podcasts and now you can turn any knowledge you want into an easily digestible audio format. The outputs are remarkably good. Businesses should be jumping on this for training, and parents should be jumping on this to help their kids. You can download the outputs to your phone and listen to them in the car, on the way to school, wherever... One of the best practical use cases of AI to date.
> Listen to a Sample
SALES FORCE IS "HARD PIVOTTING" TO AI AGENTS
After a recent blog post by venture capital firm Andreessen Horowitz entitled "Death of a Salesforce: Why AI Will Transform the Next Generation of Sales Tech" that stated "AI will so fundamentally reimagine the core system of record and the sales workflows that no incumbent is safe.", Salesforce CEO Marc Benioff has seemingly responded by stating he recognises the challenge and that is why Salesforce is pivoting the whole company to AI agents.
SO WHAT: I am on the same bandwagon as the Venture Cap mob - AI is going to tip traditional software development on its head and the big incumbents are at risk. But Salesforce does have one big advantage, data incumbency. They have all their customer's data in their systems so it will be easier for companies to just stick with them - if their AI tools can cut the mustard (which they don't at the moment).
> Read More
NEWSCORP BANS AI TOOLS CITING ETHICS - NO ONE BELIEVES THEM
NewsCorp has banned journalists from using an AI based meeting notetaking app. As part of the communications sent to the team to justify the ban, their CTO stated that they are "setting the highest standard for ethical use of AI in our business".
SO WHAT: See the article from a few weeks back where the NewsCorp executive chair accidently let slip they generate 3,000 articles per month using AI. When it comes to ethics I would have thought notetaking apps that help human staff were far more palatable on the ethics front than replacing human journalists with AI - but obviously I've missed something...
> Read More
AI SHORTS -
Former Google CEO and Chairman says generative AI "will usher in profound changes to the world in the next two years, at a scale that almost no one understands" - more
-
SunCorp has built its own Generative AI engine to assist staff with customer service, cybersecurity, operational efficiency, compliance, risk management, and market research - more
-
Meta is going to use Australian's Facebook and Instagram content to train its AI models and our government is doing nothing to stop them - more
-
Adobe has previewed its move into the generative AI video space, touting not only text to video but other AI tools to be added to their existing video editing software suite (its all "coming soon" apparently) - more
-
Transurban, the toll road operator, has used AI to automate the process of identifying vehicles without an e-tag, reducing human effort by 40% - more
-
Amazon is going to use AI to generate its audiobooks moving forward, using a model trained on voice actors - more
-
There are black market LLMs that you can buy and use to do bad stuff, like generate spam or support hacking - more
-
A new snazzy tool from Google will let you search your photos using Generative AI, albeit this does mean an AI model is looking at your photos, it is early access with opt in for now - more
-
Canva partially blames its up to 500% increase in subscription costs on its move to generative AI, which isn't justifiable based on the quality of their AI tools to date - more
-
Younger Aussies are apparently turning to generative AI to help them with money decisions and budgeting (probably better than nothing) - more
-
BP has signed a five-year deal with Palantir to use AI for faster decision-making in its engineering operations - more
|
KNOW SOMEONE WHO MIGHT WANT TO SUBSCRIBE? PLEASE SEND THEM THIS LINK...
> CLICK HERE |
That's it for this week. For those asking, yes, all the photos of me in these emails are not really me and are AI generated likenesses. If you have any tips on improvements or content to include then feel free to hit me up - adam.barty@revium.com.au.
|
|
|
From the desk of Adam Barty on behalf of Revium Pty Ltd Level 5, 84 Cubitt Street, Cremorne 3121 Unsubscribe |
|
|
|