The H.A.I.R raising one

Greetings, H.A.I.R readers.

We're going to cover a few things today:

• Recap of the AI Safety Summit
• Recap of OpenAI's Dev Day announcements
• X.ai's new Twitter Chatbot
• Another book recommendation
• Events
• AI Training

So let's get on with it, shall we?

Agreement Reached on AI Testing at the AI Safety Summit

If you’ve somehow managed to miss all summit coverage from last week, here’s a summary in brief: Leaders from around the world convened at Bletchley Park, England, to discuss safe AI regulation with the US, UK, EU & and China already reaching an agreement to collectively manage the risk of AI.

Here are the key takeaways to ignite a conversation:

• The Bletchley Declaration, signed by the EU and 28 countries, commits to creating safe and human-centric AI.
• Attendees aim to build a shared, evidence-based understanding of AI risks to inform future policies.
• The event's main focus is mitigating the risk of malicious use of advanced AI systems.
• Some, including Meta's Yann LeCun and Google Brain's Co-Founder Andrew Ng, argue that emphasising AI's existential threat could stifle open-source efforts and innovation.
• Surprisingly, China participated with delegates from its Ministry of Science and Technology, Alibaba, and Tencent, signalling global collaboration.
• The Bletchley Park event marked the first global AI safety summit, with plans for future events in South Korea and France in 2024.

All the Announcements From OpenAI's DevDay

So there we have it. Yesterday saw OpenAI's first DevDay, which came almost a year after the launch of ChatGPT. Since that launch, we've seen a slew of product announcements as OpenAI strives to maintain its early lead in the Generative AI space

So, what was announced at the first DevDay:

• OpenAI is launching a GPT store later this month. The store will let users share and sell their custom GPT bots. Sam Altman says OpenAI is going to “pay people who build the most useful and the most used GPTs” a portion of the company’s revenue.
• ChatGPT is getting GPT-4 Turbo. ChatGPT uses the latest version of OpenAI’s large language model, allowing it to browse the web to write and run code, analyze data, and more. It also won’t have the “annoying” drop-down model-picker menu — it will know which model to use automatically. GPT-4 Turbo, the latest version of the company’s GPT large language model. It comes with several improvements, including “better world knowledge” and a 128,000 token window, allowing for longer prompts.
• You’ll be able to make a custom ChatGPT bot. They’re called GPTs and are meant to be tailored to specific uses. You’ll be able to add custom instructions, knowledge, and actions and can program them by typing what you want them to do.
• One hundred million people are using ChatGPT on a weekly basis. OpenAI CEO Sam Altman announced user numbers. Since releasing its ChatGPT and Whisper models via API in March, the company also now boasts over two million developers, including over 92 percent of Fortune 500 companies.

Grok - X.ai's Chatbot for Twitter

This weekend X.ai introduced its first product for select X (formerly known as Twitter) users. The product is called Grok. It is a chatbot with direct parallels to ChatGPT. A key difference is that Grok has real-time access to X data, which presumably offers up-to-date and unique information for users.

Grok is an AI modeled after the Hitchhiker’s Guide to the Galaxy, so intended to answer almost anything and, far harder, even suggest what questions to ask!

Grok is designed to answer questions with a bit of wit and has a rebellious streak, so please don’t use it if you hate humor!

A unique and fundamental advantage of Grok is that it has real-time knowledge of the world via the 𝕏 platform. It will also answer spicy questions that are rejected by most other AI systems.

Grok is still a very early beta product – the best we could do with 2 months of training – so expect it to improve rapidly with each passing week with your help.

Thank you, the xAI Team

Musk Refers to Grok as an AI Assistant

Information on the new solution is still limited, but it is arriving quickly as some early access beta testers begin to publish their Grok results. Examples of real-time information sharing, answering technical questions, and writing Python code have emerged.

There is a sign-up page for the waiting list, and Musk indicated it will be available exclusively to Twitter Premium+ subscribers. Premium+ is a new subscription tier that will cost $16 per month for U.S. users through the web or $22 if purchased through the iOS App Store or Google Play. Grok is initially limited to U.S. users.

Grok uses a new LLM (Large Language Model)

Grok-0 is a 33B parameter large language model (LLM) and the predecessor of Grok-1. It is not specifically stated whether Grok-1 is also a 33B parameter model. Most people commenting on Grok today assume it is, but with the improved performance between the models, it may be larger. Grok-1 has an 8k data token context window (~ 6,000 words), double the original GPT-4 implementation in ChatGPT, but likely comparable to the current interaction of OpenAI’s chatbot. The model was trained on data “up to Q3 2023,” which likely means through the end of June 2023 or shortly thereafter.

Grok is better than GPT3.5 and Llama

According to the X.ai announcement:

On these benchmarks, Grok-1 displayed strong results, surpassing all other models in its compute class, including ChatGPT-3.5 and Inflection-1. It is only surpassed by models that were trained with a significantly larger amount of training data and compute resources like GPT-4. This showcases the rapid progress we are making at xAI in training LLMs with exceptional efficiency.

Since these benchmarks can be found on the web and we can’t rule out that our models were inadvertently trained on them, we hand-graded our model (and also Claude-2 and GPT-4) on the 2023 Hungarian national high school finals in mathematics, which was published at the end of May, after we collected our dataset. Grok passed the exam with a C (59%), while Claude-2 achieved the same grade (55%), and GPT-4 got a B with 68%. All models were evaluated at temperature 0.1 and the same prompt. It must be noted that we made no effort to tune for this evaluation. This experiment served as a “real-life” test on a dataset our model was never explicitly tuned for.

Grok has a personality

Grok is not multimodal

A key Grok gap compared to ChatGPT and Google Bard is the lack of multimodal capabilities. The company says it will introduce these features in the future.

Currently, Grok doesn’t have other senses, such as vision and audio. To better assist users, we will equip Grok with these different senses that can enable broader applications, including real-time interactions and assistance.

H.A.I.R Book Club

This week's recommendation is Ethical Machines by Reid Blackma

n

If you’re interested in deploying AI in your organisation, this is a great read. It’s always at the front of my Kindle library as I just keep going back to it because it gives so many insightful overviews of the different things to consider if you’re integrating AI into your organisation.

Events

This week sees me joining the team at Oleeo, where I'll be speaking at their customer conference on the use of AI in Recruitment. Their customers will be getting a hands-on demo of some of the cool things that they can do with ChatGPT and then we'll be having a fireside chat on some of the more interesting sides of implementing AI into your recruitment process

Last week saw me presenting the Keynote at Recruitment Tech's #RTE23 event on why Conversations with Machines will be the future skill that recruiters need.

AI Training

I don't normally take the opportunity to self-advertise in this newsletter, but I wanted to share the latest feedback that I received from a recent Generative AI training session that I ran for a Global TA Team.

Sponsor H.A.I.R - Putting the AI in HR

Are you ready to shine in the world of AI-driven HR and recruitment? Have a product that you want to show off? Join us as a proud sponsor of H.A.I.R, the go-to newsletter for all things conversational and generative AI in HR.

Why Sponsor H.A.I.R?

🚀 Reach a targeted audience: Showcase your solutions to HR professionals, recruiters, and industry leaders passionate about AI.

🤝 Build lasting partnerships: Collaborate with thought leaders in AI and HR, forging valuable connections.

📈 Boost your brand: Elevate your AI offerings and position your company as an industry innovator.

📣 Amplify your message: Gain exclusive visibility in a newsletter dedicated to HR AI insights.

Contact me today to explore sponsorship opportunities.

I'm Martyn Redstone - founder of PPLBOTS, AI training and consulting for Recruitment and HR Leaders, and founder of Bot Jobs, the leading job board for Conversational AI roles globally.

I'm always here to chat about all things HR, AI, and the future of recruitment. If you have questions, insights to share, or just want to connect, drop me a line. Let's continue this conversation!

Martyn