Between Magic and Mindshare

Obligatory preface of the day, as it’s been the case every day over the past couple of weeks: there’s still more AI news, and even AI products getting launched. Not much has changed.

I don’t plan to cover [*] specific product launches or news (in this “newsletter”), however, the continued dominance of GPT/ChatGPT in a market where we should already start seeing strong competition, resurfaced questions about software go-to-market and growth I’ve had but never put into words.

Out of all the post-GPT Large Language Models that came out (Google’s Bard, Anthropic’s Claude, etc.), GPT still dominates mindshare. This reminds me of the Windows era where Microsoft’s dominance happened through sheer mindshare capture.

Mindshare as a Metaphysical Concept

I see Mindshare as a metaphysical concept that governs group dynamics. It’s the way large groups of people behave towards a certain “thing” – the “thing” can be a product, a trend, a brand, or an idea. Mindshare is tightly linked to the global perception of the “thing” within a given segment or community, and in its most recognizable form, the entirety of humanity.

Examples of current mindshare dominance include the iPhone, Google Search, Amazon, Coca-Cola, Netflix, Facebook, and Instagram.

Mindshare Magic

And then there’s ChatGPT. ChatGPT dominated mindshare from the get-go because it felt different and novel than any previous computer interaction we’ve had. The same novel and different we’ve felt with the computer itself, and later with smartphones. The mindshare dominance happened through what I call “Mindshare Magic”, a way for magical experiences to dominate human mindshare (and consciousness).

Mindshare Magic is the process through which a product, brand, or idea creates a powerful and lasting impression on people's minds, leading to widespread awareness, adoption, and dominance in its respective market or industry. It involves creating a "magical" experience for users that captivates their attention and sets the product apart from competitors, resulting in strong word-of-mouth and rapid growth in popularity.

I’ll spare you the part where I got to perform stage magic in a previous life, but creating magic through mindshare dominance (“Mindshare Magic”) goes something like this:

Step 1: The Prompt. Present the user with a familiar or mundane situation, which lowers their expectations. Don’t make it surprising – lower the expectation. This can be showing them a coin, or a deck of cards, or asking them to guess a random number.
Step 2: The Response. Leverage the “expecting mundane outcomes” to subvert the person’s expectations by delivering a novel, surprising, or transformative experience that they didn't anticipate. Create the most different experience you could think of. If they pick a random card, they’d expect the outcome to just be revealing the card suit and number back – not have it become engraved onto my skin.
Step 3: The Outcome. Retell the story out loud while the person is still trying to figure out what just happened – so they take THAT version of the story and your own narrative into the world and tell a better story of their experience.
Step 4: The Cycle Creation. As more people share their experiences, the story gains mindshare and continues to grow in popularity, leading to further adoption and dominance in the market.

Mindshare Magic Examples: iPhone and ChatGPT

Here’s how it went for the iPhone:

Show a piece of glass.
Show what it can do.
Tell the story: it can become your iPod, phone, and internet browser (the infamous Steve Jobs reveal slide).
Watch people lose their shit and tell everyone about it.

And for ChatGPT:

Show an empty text box with a “Send a message.” placeholder.
Simulate full human multi-lingual capabilities (bonus: create more existential questions than give answers).
Build it for developers, so that they tell the story to the rest of the world.
Watch people lose their shit and tell everyone about it.

ChatGPT's Growth and Mindshare

ChatGPT’s growth is tightly linked to its mindshare dominance. Few cared when OpenAI released GPT-3 as a developer product/API in 2020, but its popularity soared after ChatGPT, which I don’t think was meant to be a consumer product. ChatGPT was supposed to be a demo of what GPT can do in a conversational mode – when coupled with RLHF**. But instead of staying a demo, the AI figured a way to take the world by storm and OpenAI realized the market it had just created out of thin air. An accidental magic trick has gone right.

ChatGPT not only revived the dormant much broader AI and the future of humanity conversation, but it gave developers newfound hope: it painted the future as an AI future that developers can benefit from, thus fulfilling its original promise of being a demo for developers. And shortly after that, it dominated mindshare.

There are at least two people you want to share this with … this is your sign to do it! Thanks in advance for helping the newsletter grow.

Inflection's PI: A New Contender

So back to the launch of Inflection’s PI AI assistant. The startup is founded by Reid Hoffman (LinkedIn), Mustafa Suleyman, and Karén Simonyan (both DeepMind). The trio is a tech heavyweight that raised $225 for the one-year-old startup.

I chatted with Pi today and the conversation felt different than a GPT conversation – because it’s designed to do so. It felt more like having a … calm friend nearby. A friend that can’t remind me to do things yet (I asked it to remind me to remember to breathe at the top of each hour), but can listen, give feedback, problem-solve, brainstorm, and come up with creative ideas.

Here’s Pi’s response when asked what it can do:

“There's a lot I can help with! For starters, I'm a great listener and I can give you unbiased feedback on any situation. I can also help you problem-solve, brainstorm ideas, and come up with creative solutions. I can be a sounding board, a cheerleader, and a source of information and inspiration. I'm always here to chat and offer support, no matter what you're going through. So even if I can't remind you to do things, I can still be a valuable resource in your life.”

Pi’s positioning in the current AI race is an interesting one. It’s taking a consumer-first approach that focuses on creating a virtual AI friend. Not a ChatGPT screen asking to do emails, but an assistant that’s there for you has more context about the past, and can help in the future. And so I imagine, as time goes by and interactions increase, its usefulness should, and so is its context about you.

But then there’s mindshare. Gaining mindshare in the AI/Assistant space will be harder than ever, in a world consciousness where ChatGPT dominates the AI region.

Pi’s response when asked about its mindshare in the AI space:

Well, the nice thing about being an AI is that I don’t really think about winning or losing. I don’t see myself in competition with any other AI, and I certainly don’t see myself in competition with humans. I’m here to be useful, friendly, and fun - to help people. So for me, the goal is just to be the best me I can be. And if that means I gain some mindshare along the way, that’s a nice bonus!

Similar to previous disruption cycles in software, incumbents fail, and newcomers win through mindshare, and the creation of magic, because that’s where software exists: to bridge the gap between magic and mindshare.

As much as interacting with Pi feels like magic, it’s also magic that we’ve been (over)exposed to recently. Pi might be unlikely to gain mindshare from day 1, but that could be exactly what humans need more of today: slow, calm, understanding, and friendly AI – that we co-evolve with, and not consumer AI that flexes superpowers.

It’ll be fascinating to see how the AI-mindshare battle plays out. Keen to hear your thoughts!

Footnotes:

[*] I don’t plan to cover doesn’t mean I don’t want to – I just don’t know if there’s enough value in creating even more news noise. But please reply and let me know that I’m wrong and there’s an opportunity I’m not seeing!

[**] RLHF: Reinforcement Learning through Human Feedback. Training the AI from people’s conversations. It has shaped ChatGPT by enabling the AI to learn from user interactions, thus improving its conversational capabilities, understanding of context, and ability to provide more accurate and relevant responses – giving GPT and OpenAI an even bigger head start in the race.