Don’t worry. You’re in the right place. We just moved things around a little.
Welcome to the new and improved A.Team newsletter: Build Mode has been brewing in our think tank for a few months now and is finally ready to serve.
If you're among the eagle-eyed readers who spotted our survey and shared your feedback, take a bow. Your insights were the secret sauce we needed to finalize our relaunch.
We heard your calls: You’re AI-obsessed, and you're hungry for more. You want to learn from case studies, insights, and expert analysis on how to 10x what your team can build in the age of AI.
With Build Mode, you’ll get access to case studies from top AI innovators plus one extra special new ingredient: a direct line to our Editorial Board's AI wisdom.
Our Editorial Board features senior tech leaders, AI mad scientists, game-changing entrepreneurs, and visionary investors. Like AJ Thomas, Chaos Pilot and Head of Talent at Google X. Every week, we’ll be tapping their insights and reactions to news, so you can have the smartest takes in the board room.
Speaking of, the AI scene has been buzzing with the kind of energy that would make a beehive look sedate. Elon Musk introduced Grok—a chatbot inspired by The Hitchhiker’s Guide to the Galaxy (rumor has it, it just keeps responding “42”). OpenAI's dev day just announced ChatGPTs, custom chatbots you can personalize—that are a boon for power users, and a blow to startups building AI chatbot "wrappers." And with the ink barely dry on Biden’s recent Executive Order, it's crunch time for anyone working with AI.
May Habib, CEO of Writer, laid out a guide for leaders trying to keep up:
- If you think that strung-together GPT digital assistants will outshine a solution that's been meticulously crafted to fit your enterprise's workflow—think again.
- Human teams are like sommeliers—they understand the nuanced flavors of low-signal scenarios. They're necessary to enhance AI and won’t be replaced by it.
- Remember that accuracy falls below 50% and hallucination rates increase by 20% once context grows.
- OpenAI's latest update on RAG is notable — not for its scalability to enterprise-grade applications, but as a testament that VectorDataBase+embeddings isn't yet the scalable, cost-effective, or swift solution for impactful digital assistant use cases in the enterprise.
CHART OF THE WEEK
LLMs Want You to Bring the Emotion
Can LLMs respond to emotional cues?
It's been a long-standing belief that the human touch remains unmatched. However, recent experiments suggest otherwise.
A group of researchers put the emotional acumen of Large Language Models under the microscope and found that LLMs may be closing in on our uniquely human traits.
By integrating what's termed an "EmotionPrompt" into LLMs — essentially combining the standard prompt with an emotional nudge — these models started showing a notable uptick in performance.
In an experiment, human participants evaluated the quality of outputs from LLMs using both standard and emotionally charged prompts. The results
Not just enhanced performance, but a significant boost in truthful and ethically responsible outputs.
FROM THE EDITORIAL BOARD
What’s Your Most Controversial AI Prediction for 2024?
From Jared Cocken, Fractional Chief Product Officer, and the guy who helped McGraw Hill build Sharpen, the TikTok for textbooks that’s become a viral sensation.
Next year will be crunch time for how AI fits into the P&L of most Fortune 500s. We’ll likely see general economic contraction in Q1, which means companies have to innovate or cut costs to make their numbers. Generative AI offers the prospect of both, but many companies will do it wrong. Some Fortune 500s will have implemented open source models due to their concern about companies like OpenAI using their proprietary data to train their models—which is a mistake. Six months in, they’ll be realizing the exorbitant on-demand cost of running, say, Llama2 on g5.48xlarge instances, and hiring and training the teams to run them. There will be a shift back to using OpenAI’s API on demand instead.
Speaking of on-demand costs, the smartest companies will realize that most people ask the same question over and over again; they’ll build a caching system, where internal experts verify the LLMs responses to work around the hallucination problem. Caching means lower on-demand fees over time.
Perceptive businesses will eschew stuffing generic chatbots into already bloated feature sets—sadly a current popular implementation—instead weaving AI into their products in ways that enhance and extend their products’ current capabilities. Just as Adobe added Generative Fill alongside Photoshop’s regular fill, we can expect writing applications to integrate AI for sophisticated text rewriting, dictionaries to provide contextually enriched ‘use it in a sentence’ examples, and educational platforms to use AI to dynamically adjust content difficulty.
Resources and Strategies for Leaders Navigating Wartime Challenges
In times of global upheaval, the question, "How can I be a better leader?" takes on a new urgency.
In the wake of the recent tragedies in Israel, our role extends beyond business as usual. We've been in close conversations with our builders, customers, and community about the challenges they’re facing and wanted to pass on some of the strategies that have been incredibly helpful for us as we all strive to move forward together
For leaders looking for direction on how to support their teams during these wartime challenges, this article dives into the pressing questions our community is wrestling with:
“How can managers lead with conviction amidst such uncertainty?”
“What are the best practices for mental health support in our teams?”
“How do we cultivate resilience while leading with empathy?”
“Is it appropriate for your organization to make a public statement?”
Armed with insights from mental health professionals we’ve worked with and actionable frameworks for navigating volatility and uncertainty, this resource is crafted to provide leaders with the tools they need to steer their teams through these challenging times with support and strength.
As we continue to support our community affected by the conflict in Israel and the ongoing struggles in Ukraine, we’re using the A.Team platform to connect non-profits, builders, and startups who either need assistance to keep projects moving or are looking to volunteer their skills and services. If you’re looking to get involved to either offer your expertise or request the support you need; sign up here.
FUTURE OF WORK
How will product management change in the age of AI?
With Matan-Paul Shetrit, A.Team’s Head of Product.
How will product management change in the age of AI?
PMs are always writing up strategy docs and decks, and LLMs give you the act of brainstorming in your pocket, which is extremely valuable. Instead of me taking 20 hours to write a strategy doc, it might take me five, right? So I can spend more time on thinking and exploring. I can ask these tools: Does the strategy make sense? Help me refine the argument. If you needed to poke holes, how would you go about it? What are the downsides? What am I missing? It's really like having a tough, skeptical investor in your pocket.
If you had to quantify, how much of a productivity gain have you gotten from these tools?
It compresses some areas of my work from 10s of hours to single digits.
As a leader, what is your AI message to your team?
Things that used to take us a lot of time don't take us a lot of time anymore. So, the excuses are over. Why is it taking so long? Why can't we use these instruments to accelerate? At the same time, I don't want someone to give me an LLM response without reading or editing. Getting from 80% quality to 100% quality is always the hardest part, as with everything in life. But instead of it taking 20 hours to get from zero to 80%, it takes me one hour. Then, overall, getting to 100 is much faster.
How should PMs be thinking about harnessing AI in their products?
The big question for me is: What experiences does this unlock for the customer? Most companies that say they’re “AI companies” aren’t actually AI companies. They’re solving a customer problem and AI is the tool they’re using. If a gardener comes in to work on your backyard, he doesn’t call his company a raking company. The rake is a tool. What we as PMs do is solve customer problems. I think that's both very humbling and beautiful.
AI DISCOVERY ZONE
Wish you had an AI Twin? D-ID just launched on the App Store, so now you can generate AI avatar videos straight from your phone.
DEEP DIVES FROM THE ARCHIVES
- Zen and the Art of Building a "Self-Managing" Team
- 3 Things Generative AI Can Never Do Better Than Humans
MEME OF THE WEEK