What Google Gemini Teaches Us About Trust and The Future

AI with ALLIE

The professional’s guide to quick AI bites for your personal life, work life, and beyond.

December 15, 2023

This newsletter is a two-parter. Strap in.

I want to walk through the highlights and lowlights of the recent Google announcement, and then I’m going to use it to extrapolate some big business insights and bold predictions for the future.

Here we go.

What Google Gemini teaches us about trust in AI

Remember the large language model buzz from the Google I/O conference this summer? Well, the tech giant finally let the cat out of the bag (a bit early, actually), and CEO Sundar Pichai gave us a look into Gemini AI.

What is it: 

Gemini AI is Google’s latest-and-greatest AI model that can “reason” across text, images, video, audio, and code — it’s multimodal.

In the new video, Gemini is seen playing a guessing game (spotting a duck from just lines and shapes), waxing poetic about the duck (can it float?), and providing opinions (a loop on a roller coaster makes it more fun).

Here is the video that was released:

The video kind of suggested a magical AI world where you could have a voice conversation with Gemini and use/transform almost any input and output.

But that world is still a wee bit away. Google is now facing criticism after people questioned the legitimacy of the video (anyone remember the Google Duplex hairstylist demo from 2018?). Within the description of the video, Google has a short statement, “For the purposes of this demo, latency has been reduced, and Gemini outputs have been shortened for brevity” but it is unclear when that caption was included. A day after the video was released, it was found that the researchers did not prompt Gemini with spoken language and video, but rather text prompts and still images. A Google Exec only confirmed it was shortened.

What we learned:

• Openness is helpful but not enough: Google published their whitepaper and actual prompts (something we don’t see often from other AI product releases these days!), but the bigger story was still the altered demo. 

• AI limitations matter: Trust needs to carry throughout your product experience. In the AI age, people want to know what your products can do and what they CANNOT do. People don’t want the magical mystery tour—they want reality. It’s better to be upfront and not perfect.

• Trust is also earned through people: Trust and transparency are not only found in the products and tech you build, but in what your executives and employees share.

What your business should do:

The takeaway is: building, maintaining, and growing your trust needs to be your #1 priority. And AI can be a tool that grows or compromises that.

We have seen multiple instances of people losing trust in organizations, companies, people, and products—like fake authors, AI misuse, or the above example of fake demos.

This is your time to be open and honest (I’ll start: I don’t clean my home often enough, I forget to call my friends back, I have to google how to spell vacuum each time, I’ve never been an engineer, I cry during commercials, I’m out of shape, and I wish I was a better debater—see how easy that was?). Be upfront about caveats and product limitations.

And it’s a time to use that to your advantage (I set up more in-person time with friends, I rarely use the word vacuum, I hired a Senior Machine Learning Engineer on my team, I keep tissues near my couch, I asked a friend to coach me on debate, and I’m checking out dance classes in NYC to explore my city). Reframe your gaps as a progress update, or call to action to recruit or ask for feedback.

I recently did an interview with Fortune (article to come soon) on how to build an AI-first company and why trust is so important in the AI age. I can’t wait to share it with you all when it’s ready.

We interrupt this newsletter to bring you…

Grab ‘The AI Daily Planner’—your year-long tech sidekick for growing your AI knowledge, organizing your daily habits, and reaching your goals. Every planner includes: 12 SURPRISE AI Tools of the Month, 60 Full-Page AI Bites of the Week, Habit Tracking and Goal Management, Allie’s Viral Goal Planning Template, Focus of the Day, AI Usage Tracker, and more.

The AI Daily Planner is available in both a spiral-bound notebook and a PDF format. It's undated too, so you can start whenever you're ready, even if it's mid-year.

This planner is perfect for anyone eager to stay on top of AI trends or for teams wanting to boost their AI knowledge. A great gift too.

…now back to your regular programming.

What Google Gemini teaches us about the future of AI

Gemini Ultra seems to be a big step forward in AI development. We’ll have to wait a bit for the big release next year where developers and product builders will get to use it, but I’m hopeful.

And we’re also seeing new releases from Mistral and Liquid AI that are changing how we think about productionizing LLMS.

This Google release, and frankly the last several months in AI, is screaming two words to me: REASONING and PRODUCTION.

1) We're moving from 'knowledge' to 'action' - AI is going to move further upstream and will be tasked with taking action. And guess what, actions require reasoning. Something I've stressed heavily for the last several years is the need for critical reasoning in this process. That's the 2024 focus—AI moving into proactive interventions.

2) We're getting more efficient - AI has become 16500x more efficient in the last 11 years, and we're starting to see that for state-of-the-art large language models. Say what you will about GPT-4 performance, but it is too expensive for most companies (including OpenAI itself!) to run at scale. Assume 2024 brings lower AI OpEx.

3) It's multi-modal from here on out - ChatGPT is multi-modal and doesn't require the user to preselect what modality they want to use, Gemini seems to be the same. There will be no lines between images, music, text, voice (always a big part of Google demos!), and video. Just one pile of "data". Assume that 2024 is multi-modal.

4) There's no one model to rule them all - we've now heard this from Andrej Karpathy, Google DeepMind, and more, but there will be different models for different use cases. Gemini has 3 tiers: Pro, Ultra, and Nano. Pro is used in Bard and Google products. Nano is used inside Google Pixel phones. Ultra is for "I have deep pockets and want the highest grade possible." Performance will be a consideration: with a score of 90.0%, Gemini Ultra is the first model to outperform human experts on MMLU (massive multitask language understanding). Assume that 2024 has more multi-model orchestration and delegation.

Gemini Pro is out and available for all to try in Google Bard.

TL;DR

1. We're moving from 'knowledge' to 'action'.

   AI moving into proactive interventions.

2. We're getting more efficient.

   Assume 2024 brings lower AI OpEx.

3. It's multi-modal from here on out.

   Assume 2024 is multi-modal.

4. There's no one model to rule them all.

   Assume 2024 has more multi-model orchestration & delegation.

Stay curious, stay informed,

Allie

Tools, courses, and blogs that caught my eye:

Over the last few weeks, I’ve pulled together some of the top releases, and my take on each one. Check it out.

1. Ramblefix - If you’ve ever had a brilliant idea but typing it out felt like a chore, now you can just record a voice memo and bam – you've got yourself a press release, social post, or even an article. It's like having a personal scribe in your pocket (try it) (my thoughts)

2. RunwayML General World Models (GWMs) - A future AI system that aims to create an internal representation of an environment, and uses it to predict future events within that environment (read it) (my thoughts)

3. Leonardo AI - Your gateway to creating visual magic with a few strokes. Leonardo AI is a live drawing canvas that uses AI to instantly create visual assets. Perfect for when you need to whip up graphics but are short on time or inspiration (try it) (my thoughts)

4. HeyGen - AI-generated avatars and voices for studio-quality video creation. It can translate into 40+ languages and has 300+ voices. I shared this a few months ago, but the product keeps crushing it (try it) (my thoughts)

5. Keywords Everywhere - ChatGPT templates that easily plug-in to the app for professionals including accountants, entrepreneurs, finance, human resources, lawyers, musicians, realtors, and teachers; can’t believe more newbies aren’t using this (download)(demo)

6. OpenAI announces new startup cohort - The OpenAI startup fund just announced Converge 2 (the second cohort of their startup incubator), a six-week program for top AI entrepreneurs. Every participant receives $1M in investment from OpenAI. That’s a nice chunk of compute! If you’re an AI founder, check it out (apply here)

Feedback is a Gift

I would love to know what you thought of this newsletter and any feedback you have for me. Did you have a favorite part? Wish I would change something? Felt confused? Please reply and share your thoughts or just take the poll below so I can continue to improve and deliver value for you all.