🚀 PR Futurist 76 |🤖 2024 the year to explore AI| 🙏 Remembering Derek | 🗳 2024 The year of elections  15 January 2024

OpenAI launches ChatGPT for teams

OpenAI has now released ChatGPT Team. It's potentially a game changer for business adoption of AI as it enables a company wide approach to AI with teams collaborating to create unique company ways of using AI.

One of the key benefits of ChatGPT Team is it solves the problem of uploading confidential data to AI and it being incorporated into its data set. Any data uploaded into ChatGPT Team remains confidential.

It's not the only team solution as at Purposeful Relations we've been experimenting with Team-GPT for sometime.

It does highlight how difficult it is for companies to decide on the right approach to AI. Should it be ChatGPT Team? Or Microsft 365 Copilot when it launches on business licences? Or Google's Duet. Or third-party solutions like Team-GPT than can use multiple models including GPT-4 and Anthropic's Claude.

There is no one correct answer, which is why we start with helping clients to identify objectives and a strategy, before jumping straight into tools.

Stuart Bruce

linkedin.com  

OpenAI launches its GPT store

Coinciding with the launch of ChatGPT Team OpenAI has also opened its GPT Store. This is where you can discover and use custom versions of ChatGPT created by users. You can use it to find useful GPTs to use yourself, or you can create your own custom GPTs and publish them in the store.

Tim Bailey

openai.com  

Are custom AI models really better than generalist ones?

Many companies and organisations are embracing AI by developing their own large language models which are trained on their own proprietary (and sometimes confidential) data. Bloomberg, McKinsey, MumsNet and the UK's Government Communication Service (GCS) are some of those taking this approach.

The Wharton School's Professor Ethan Mollick has shared some fascinating research that shows custom LLMs might not be as good as the results GPT-4 can already generate. The specific example looks at Bloomberg's LLM, but he indicates it could be part of a pattern where "the smartest generalist frontier models beat specialized models in specialized topics."

Stuart Bruce

linkedin.com  

What's the future for publishing and AI?

Since the launch of ChatGPT there has been much discussion about what it means for the future of publishing and content creators of all types. Some are alarmed at AI learning from copyright content.

There are already legal battles on different aspects of the debate, but one of the most interesting is the New York Times's action against OpenAI.

This is OpenAI's public response where it explains how it supports journalism, partners with news organisations, and believe The New York Times lawsuit is without merit.

It has been much discussed and I shared my personal thoughts in a comment on a LinkedIn post by Christopher Penn.

"I initially kept changing my mind on this. There are two issues - legal and moral. On legal I'm watching developments, as there is no clear answer. On moral I'm now firmly of the belief that it's not just fine but that it's verging on immoral to say it isn't.

Copyright protects words and images. It doesn't protect ideas. The best analogy I heard was from the CEO of The Atlantic. People can go into a library and read every back issue of The Atlantic. They can even copy parts of it. That's actually something most writers and publishers would want to happen.

That's all AI is doing. It's just doing it 'better' and at scale.

Assuming the person uploading it has already paid for it then morally it is no different to them letting a friend read it or giving it to a friend to own.

There is an urgent and important need to find new ways to financially compensate creators, but stopping people or AI reading or using their work isn't it.

Not allowing AI to use copyright content is as immoral as Luddites smashing machines in the early days of the industrial revolution."

As this is such a controversial and much debated issue we'd love to know what you think.

Stuart Bruce

openai.com  

Meta's Threads will integrate with the Fediverse

The full article we've shared is long and a bit geeky, but it's significant so here is the TL;DR.

Meta is integrating its new Threads social network with the Fediverse. The Fediverse is a collection of decentralized social networks that interoperate with each other. Users of different platforms can communicate and follow each other across the Fediverse, without being locked into one company’s service. Other platforms that are connected to the Fediverse (or will be) are Mastodon and Bluesky.

This is a new approach for Meta as rather than the 'walled garden' social network approach of Facebook this is a more open one.

Potentially this means asking if Threads, Bluesky or Mastodon will 'replace' Twitter might be the wrong question. If they all interoperate then people will use the platform, or platforms, they are most comfortable with. It is early days so we have to watch how the Fediverse implementation develops on different platforms.

Stuart Bruce

plasticbag.org