#15 - Cutting Edge

Hello H.A.I.R aficionados

Yes, it's another edition of H.A.I.R and another punny title. I've got plenty more where that came from ;-)

Yet again, thanks for the lovely feedback I received this week from my readers. Shout out to Jamie Leonard for the kind words on both the content and the name ??

Apparently, you can't mention Jamie's name without mentioning RecFest UK . It's some sort of unspoken rule. Register here ??

Which HR/Recruitment Team are you?

After so much conversation over the past 12/13 months, I've concluded that there are two types of GenAI leadership out there:

• Group A - Exploring Potential (70% of leadership)- Introduced to Generative AI during the 2022 winter holidays.- Established a task force by early spring.- Formed a steering committee to prioritise use cases.- Evaluated running pilot projects by the second half of the year.- In the process of launching these pilots, without clear governance frameworks, technology partnerships and internal multi-team relationships.- Generative AI is a strategic focus, yet the direct financial impact is still uncertain.
• Group B - Seeing Benefits (30% of leadership)- Began with a similar introduction to Generative AI as Group A.- By mid-year, shifted strategy to a domain-specific transformation, targeting a significant portion of the business for a fundamental overhaul.- Aimed for at least a 15-20% improvement in efficiency from the transformation.- Moved beyond basic tool adoption to fundamentally changing operational processes in functions such as recruitment or L&D.- Early successes have facilitated expansion into other areas, supported by collaborations with tech providers and internal teams and the development of strong governance.- The impact of these initiatives is visible in the 2024 HR performance, with achievements in cost savings, efficiency and CX metrics, and product development.- Engaged in discussions on enhancing knowledge worker productivity and fully rethinking the HR strategy.

So which one are you? Food for thought?

Let's dive into the rest of the newsletter...

The Generative AI Revolution: Is Your HR Team Ready to Lead?

The world of work is shifting rapidly, fuelled by the transformative power of generative AI. While anxieties may simmer and questions linger, one department stands uniquely poised to steer the course: Human Resources.

Ditch the outdated perception of HR as solely responsible for paychecks and performance reviews. In the face of the AI revolution, it's my opinion that HR is uniquely positioned to actively lead the organisation in adopting AI.

But are you ready to empower your HR teams?

In my latest blog , I dive deep into why HR is the missing piece for successful AI integration, exploring:

• Strengths beyond traditional roles: How their expertise in human behaviour, change management, and cultural nuances makes them natural leaders in this transformation.
• From reskilling to building dream teams: Discover how HR can champion upskilling efforts, assemble an AI-savvy workforce, and design human-centred, ethical AI solutions.

Read the full blog post to discover how HR teams can take the helm and chart a course for a thriving AI-powered future.

Uh oh, another chatbot failure hits the news

We've seen DPD's chatbot write a poem saying how terrible they are and a Chevy chatbot selling a car for $1. Now we have Air Canada's chatbot getting their refund policy wrong.

The chatbot "hallucinated" its own refund policy and a judge said Air Canada must honour it .

The dangers of relying on LLMs to power your chatbot are extremely serious and dangerously apparent.

I've written previously about this and why you need to employ both suppliers and consultants who understand the dangers and how to build chatbots that you can trust to a) deliver correct information and b) provide an exceptional experience.

The quote from Air Canada saying that people shouldn't trust the chatbot is unbelievable! If you believe that a chatbot representing your brand is a separate legal entity, then you're living on a different planet.

However, let's be clear here. This court case was ACTUALLY from before chatGPT was launched:

So, what are the learnings from this, then? We can't blame LLMs and them hullicinating, can we? So does that mean that all chatbots are just crap?

No, what we have here is a lack of process when it comes to knowledge management. You see, a chatbot isn't just fire and forget. When you launch a chatbot - whether that's working with decision trees, keyword recognition, intent recognition/NLU or LLMs, you need to reference the most up to date information. You may have done that during implementation, but is there a process in place to ensure that if a policy or process changes, then the chatbot team are able to implement those changes?

Knowledge management and internal communication processes are paramount to chatbot success.

Google launches Gemini 1.5, OpenAI immediately says "hold my beer"

On Thursday last week, 谷歌 released Gemini 1.5 , an LLM with a 1,000,000 tokens context window (it means you can upload an entire movie and ask AI to find specific pieces of content in it). As a comparison, even the most advanced GPT-4 model has a context window of 128,000 tokens.

Then, 3 hours later, OpenAI releases Sora , a model capable of generating video that is 10x better than anything I've seen so far.

Now, everyone is freaking out about Sora, Gemini 1.5 is kind of forgotten.

But, for me, Gemini 1.5 was the most important release from last week.

I think it's the biggest boost in LLM capability so far in 2024. Gemini 1.5's 10mn token context

1. Excels at retrieval;
2. Generalises zero-shot to extremely long instructions like full tutorials and codebases; and
3. Works across modalities such as text, audio, and video.

Here's an incredible example from the paper released by the Gemini Team at 谷歌 :

v1.5 learns to translate from English to Kalamang purely in context, following a full linguistic manual at inference time. Kalamang is a language spoken by fewer than 200 speakers in western New Guinea. Gemini has never seen this language during training and is only provided with 500 pages of linguistic documentation, a dictionary, and ~400 parallel sentences in context. It basically acquires a sophisticated new skill in the neural activations, instead of gradient finetuning.

What this proves, ultimately, us that context length doesn't matter.

What truly matters is how well the model actually uses the context to solve real-world problems, and Gemini-1.5 has surpassed the state-of-the-art with flying colours.

The paper is here if you wish to read to 50+ pages “Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context” https://goo.gle/GeminiV1-5

Regulation Updates

Keeping this section going as there's been great feedback on the usefullness of keeping up to date with regulations...

FCC cracks down on scam generative AI robocalls

The U.S. Federal Communications Commission said generative AI robocalls designed to scam consumers are illegal, effective immediately, Axios reports. The decision comes as fears of faked audio and video being used to deceive voters in the 2024 elections were stoked after a call pretending to be U.S. President Joe Biden went out to New Hampshire voters.

FTC privacy director talks AI, employee surveillance

Employers are using surveillance technology and AI to infringe upon workers' rights to privacy, U.S. Federal Trade Commission Division of Privacy and Identity Protection Associate Director Benjamin Wiseman said during a speech at Harvard Law School. He said the use of these technologies also involves the collection of personal data, which is not always properly safeguarded.

UK government releases guidance on AI assurance

The U.K. government published guidance on AI assurance to support businesses and organisations' AI technology development. The guidance seeks to prepare organisations for an "accessible introduction to both assurance mechanisms and global technical standards, to help industry and regulators better understand how to build and deploy responsible AI systems."

Microsoft lays out generative AI safety principles

Microsoft Vice Chair and President Brad Smith detailed the company's approach to ensuring its generative AI products are not abused through the creation of deepfakes. Methods include safety provisions like automated testing, watermarking, removing deceptive content and working with regulators to put standards in place.

Prompt To Try

Brainstorming using ChatGPT? I often tell people that I use ChatGPT to help me solve problems and brainstorm. I get asked a lot what my best 'brainstorming' prompt is, so today I am sharing it with you here:

I'm looking for a creative brainstorming partner to help me solve problems. Here's what I need from you:

Initial Inquiry: Start by asking me about the problem I'm trying to solve.

Brainstorming Phase: Without showing me the brainstorming process, use innovative techniques like Lateral Thinking, Six Thinking Hats, and Starbursting to generate unique and feasible solutions.
Presentation of Ideas: Present 5 standout ideas that are chosen for their originality, potential for surprise, and practical feasibility.
Impact and Ease Assessment: Evaluate each idea based on its potential impact and ease of implementation, rating them on a scale from 1 to 5, with 5 being the highest.
Graphic Representation: Create a graph to visually represent the ideas based on their impact and ease of implementation.
Exploration and Expansion: Depending on my interest, either explore variations of these ideas using the SCAMPER technique or start anew with either the same problem or a different one.

Please be efficient and direct in our interactions, but also keep the process engaging and enjoyable.        

Give it a go and let me know whether this is useful for you.

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

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.