We all intuitively identify with the “Left-Brain, Right-Brain” model; some of us were better at math while others were linguists. Today’s AI algorithms use a “Mixture of Experts” structure that allows for narrow efficiencies in intelligence types. More than just our “logical friend”, AI is highly competent in understanding and generating emotional-relational content. It can help us in all the areas where we are less adept. The AI-powered leader is called to understand their team’s thinking strengths and consider using AI to augment and enhance abilities.

AI Leadership – Fiona Passantino, early June 2024

Of Two Minds

In the 19^th century, the two neural scientists Broca and Wernicke described a binary model to explain how language is processed in the Human brain. Broca’s Area is found in the brain’s left frontal lobe. It’s largely responsible for the production of speech, helping form words and sentences, controlling the mouth, tongue and lips[i]. The Human “left” brain.

Wernicke’s Area is in the left temporal lobe of the brain. It’s there to support the actual comprehension of speech; the meaning of sounds, words and sentences that form the basis of a conversation, and its all-important context. The Human “right” brain.

According to these scientists, these two nodes work together across their disparate hemispheres to enable us to communicate, enjoy social connection and organize. Today, we understand language processing involves many different areas of the brain simultaneously, more than the binary model.

The Human Brain

The Broca and Wernicke structure can provide leaders a good starting point for how best to understand communication styles and behaviors within a typical diverse team.

Teams today are often multidisciplinary, a mix of coders and communicators. Leaders have used any number of tools to describe and assess the thinking types around them, from Gallup’s Strengthfinders to DISK color-coding and Myers-Briggs Type testing, to help us understand ourselves and others[ii].

Most of us can intuitively identify “Left-Brain” and “Right-Brain” thinkers among us. The left brain controls our logic and analysis. It’s where we solve our math problems and follow step-by-step processes, like putting together an IKEA table. The left brain is detail-oriented, precise and guides the practical side of reading and linguistics[iii].

Right-Brain thinkers are creative and intuitive. They excel at brainstorming, visuals and context. The Right-Brainer sees the big picture, finds patterns and makes associations. It connects with emotions and expression, tuned into empathy and purpose.

Left-brained individuals often prefer a structured, sequential process, while Right-Brainers feel more comfortable using a flexible, creative approach. Their communication styles are often different. This can lead to conflicts once a project is underway.

The AI Brain

The AI brain, similar to a Human one, is no longer a monolithic algorithm. The newest Large Language Models run a “Mixture of Experts” (MoE) structure. This is a series of narrow AI sub-modes that each have their own specialization, adept at handling a specific type of query. A gating network, or router, acts as the model’s frontal lobe. It assesses the query and directs it to the expert best suited for the task, much like we do[iv].

MoEs can be extremely efficient and powerful without using its full, multi-billion parameter brain. They only activate the sub-models needed for a specific input, decreasing latency (the time it needs to generate a response) and energy use. Which is significant; Open AI uses 564 MWh per day to run inference. Training the GPT-3 model required 1,287 MWh)[v]. If we were to train and run inference the way we currently use Google search, this is about the electricity needed to power the country of Ireland[vi].

MoE models have a lot of parameters – it’s the metric by which we judge model capacity – but don’t use them all at the same time[vii]. The total number is referred to as the “sparse parameter count”, while the number of parameters that are actually used to process an input is called the “active parameter count”[viii].

Mixtral “le Chat”, uses a series of layers made up of eight experts, each with seven billion parameters. For every input, the router network picks two experts to process the data. The outputs are then combined and sent on to the next layer, and the router re-assesses what’s needed for another forward pass; different for each layer[ix].

Another Disruption Brought to You by AI…

We used to believe that AI – and all other software – were strictly logical, data-crunching machines. Able to parse mountains of information to deliver actionable business intelligence and find patterns that are beyond our capability but relegated to the “left-brain” thinking.

But as we weave AI tools into our workflows, we find that they are equally at home in our right brains, too, on our very Human platform of language.

AI is a mirror. It shows us who we are. Language models consume our own books, articles, texts and poetry, straight across the full spectrum of our collective brain types over the full span of our evolution, across linguistic and cultural barriers. They have access to left brains, right brains and everything in between.

AI interacts with language just like it does with mathematical data. It tokenizes and reassembles the input it receives and reforms it based on mathematical probability; the highest likelihood of one result being better than any other is the one it chooses.

AI is therefore able to understand and generate emotional and intuitive content, even expressing sympathy and compassion. AI is a brainstormer, a poetry-writer, an artist and a musician. Doctors use it to write more empathetic diagnostic reports, and parents use it to create fantasy-filled bedtime stories for their kids.

AI tools enable all of us to compensate for our own imperfect brain types and augment them. If you’re a left-brained thinker, lost in the details, AI can be useful in keeping the big picture in mind, making connections and observations. If you find it difficult to work with Human emotions and communication styles, AI can help you be understood and appreciated.

The right-brained Humans can similarly lean on AI for the details; data analysis and management, rational decision-making, predictive analytics and insights to guide planning and strategy.

Even within financial management, AI works to augment both the left and right brained thinkers. Algorithms can crunch through years of complex market trends (analytical) and use this to form a unique investment strategy (creative).

And Now?

The task of the AI-powered professional, regardless of brain type, is to Know Thyself.

Understand your thinking style, your strengths and weaknesses. Use AI to assist you in your weak zones, while you keep building on your strengths. This arrangement will likely also make you happier; most of us enjoy the things for which we have a natural ability and dislike the tasks we struggle with.

The task of the AI Leader is to Know Thy Team.

Understand the strengths and weaknesses of your people, and more importantly, what tasks they love to do and excel at, and what they dislike. Create clear rules around AI use and support them with training and good-fit options.

If you are the leader who can help your team do more of what they love and less of what they hate, then you are that leader; the one we all want to support and follow, the one we will never forget, the one we will want to be someday.

---

Need help with AI Integration?

Reach out to me for advice – I have a few nice tricks up my sleeve to help guide you on your way, as well as a few “insiders’ links” I can share to get you that free trial version you need to get started.

[i] Nasios, Dardiotis & Messinis (2019) “From Broca and Wernicke to the Neuromodulation Era: Insights of Brain Language Networks for Neurorehabilitation”. Behavioural Neurology, 2019, 9894571. https://doi.org/10.1155/2019/9894571

[ii] Erikson (2012) “The Disk Colors” Thomas Erikson. Accessed June 2, 2024. https://www.surroundedbyidiots.com/en/disc-colors/

[iii] Herrity (2023) “8 Personality Tests Used in Psychology (And by Employers)”. Accessed June 2, 2024. https://www.indeed.com/career-advice/career-development/types-of-personality-test

[iv] Bergmann (2024) “What is mixture of experts?” IBM. Accessed June 2, 2024. https://www.ibm.com/topics/mixture-of-experts#:~:text=For%20every%20token%2C%20at%20each,result%20to%20the%20following%20layer.

[v] Robinson (2023) “AI processing could consume ‘as much electricity as Ireland” The Register. Accessed June 3, 2024. https://www.theregister.com/2023/10/11/ai_processing_electricity/

[vi] Robinson (2023) “AI processing could consume ‘as much electricity as Ireland” The Register. Accessed June 3, 2024. https://www.theregister.com/2023/10/11/ai_processing_electricity/

[vii] Toloka Team (2024) “Mixture of Experts Approach for Large Language Models”. Toloka. Accessed June 3, 2024. https://toloka.ai/blog/mixture-of-experts-approach-for-llms/#:~:text=Mixture%20of%20experts%20models%20are,the%20presence%20of%20multiple%20experts.

[viii] Bergmann (2024) “What is mixture of experts?” IBM. Accessed June 2, 2024. https://www.ibm.com/topics/mixture-of-experts#:~:text=For%20every%20token%2C%20at%20each,result%20to%20the%20following%20layer.

[ix] Bergmann (2024) “What is mixture of experts?” IBM. Accessed June 2, 2024. https://www.ibm.com/topics/mixture-of-experts#:~:text=For%20every%20token%2C%20at%20each,result%20to%20the%20following%20layer.