LOL my AI cracked me up [Horizon 01 - Issue #8]

March 17 - March 23, 2025

I know, I know. Last week, we were punching down on creativity and told you that the recent trend in AI is not about image, text, or others - but emotions and creativity. Now, AI is becoming funnier, more engaging, and more drama-lama.

And that says a lot. Any smartass human can recite academic papers. It requires basic intelligence, a hardened skin on our butts, and even more reading. But humor? It’s complex. It requires self-reflective intelligence, cultural versatility, and high emotional intelligence. Being funny is a very human thing.

Or, it was.

If AI can be as funny as most humans, it means three things:

1. AI can become not just a “tool” for teams to use for better productivity. It can become a proper “team member” who is an integrated part of the team.

2. AI can craft narratives and relatable ideas and is great at storytelling. All it takes is good instructions from you—like you would brief a human teammate.

3. Humans are still required as directors. You are the one who sparks and coins up the show. But it’s not you anymore who performs or runs it.

Harsh? Nah. Scifi? Not really. I bet this will be widespread by this year or next year.

Enjoy this week’s Horizon 01. Most of the stuff we bring you this week proves that AI can crack jokes, hunt down memes, and act as a full-time creative partner for your marketing team.

Peter & Torsten

AI Memes Better Than Humans

An AI just beat us at our own funny game.

In a new study, researchers pitted large language models against people in a “meme Turing test” – generating captions for popular meme images – and had online judges score them. The result? On average, AI-generated meme captions were rated more creative, humorous, and shareable than human-made ones. Even memes co-created by humans and AI scored lower than the AI-only memes. (Humans still produced some of the best captions, but too few to swing the averages.)

Given the proper training (in this case, OpenAI’s GPT-4o model), AI can master the art of internet humor – complete with timing, cultural references, and puns – well enough to make people prefer it over human witticisms.

If AI can win at humor, it can likely tackle other creative domains once thought exclusive to human talent. Marketers who rely on clever copy and viral memes might soon find AI to be a surprisingly adept brainstorming partner—or even competitor.

We’re already seeing brands use AI to generate social media posts; this study suggests those posts can be genuinely funny and engaging, not just boilerplate.

In the short term, creative teams can leverage AI for a first draft of ideas (taglines, ads, memes) and then add a human touch to the top contenders.

In the medium term, as AI humor and cultural savvy keep improving, marketers might orchestrate entire campaigns in which AI drafts dozens of edgy memes or jingles, and human creatives curate the best.

The takeaway

Don’t underestimate AI’s creative spark. It’s passing creativity tests, so savvy teams will use that to their advantage – freeing humans to focus on higher-level narrative and strategy while the bots riff and iterate on content.

Even More Creativity: 3D Creation and Image-to-Video

Last week, Roblox unveiled “Cube,” a generative AI system that creates 3D objects (and soon whole scenes) from text prompts. It uses a novel 3D tokenization approach to build detailed models natively. This dramatically speeds up asset production and lowers the skill barrier – even non-designers can type a description and get a 3D model.

Tencent likewise open-sourced its Hunyuan3D 2.0 suite, five AI models that turn text or images into high-quality 3D visuals in as little as 30 seconds. Tencent claims the models outperform industry peers in accuracy and visual quality.

Also last week, Stability AI released Stable Virtual Camera, an AI model that transforms 2D images into “immersive” 3D-like videos. Feed it one or more photos (up to 32), and the model will generate new views of the scene with realistic depth and perspective. Users can script complex camera movements—pan, zoom, even a spiraling dolly shot—and the model will render a smooth video following that path. Currently, a research preview (non-commercial) available on Hugging Face can produce clips of up to 1,000 frames.

How can you use these tools

Marketers could now instantly generate 3D product mockups or virtual event props, while creative teams prototype AR/VR content without 3D modeling expertise. These tools hint at an era in which immersive ads and product demos can be generated on demand.

Creative professionals can quickly turn a single product photo into a dynamic video ad, or filmmakers can pre-visualize scenes by “filming” a concept art image.

Emotional By Design

This week, a fascinating development was that a study found that people often perceive AI responses as more compassionate and understanding than those from human mental health experts.

In trials where participants sought advice or counseling, third-party assessors consistently rated the AI’s replies as 16% more compassionate on average. They preferred the AI 68% of the time over trained counselors. This held even when participants knew whether the response was from a human or a bot.

It’s a striking result – AI, lacking human emotion, somehow delivered a tone of empathy that resonated better with people. Why? Researchers suggest that AI’s training on large swaths of supportive language (and perhaps lack of fatigue or bias) allows it to respond in a calm, caring manner every time. Essentially, the chatbot excelled at “active listening” and validation, never getting frustrated or rushed as a human might.

Indeed, AI providers are actively engineering more emotional intelligence into their tools. OpenAI’s latest voice model, for instance, can be directed to speak in a certain emotional style, like a gentle, reassuring customer service agent. This means that not only can AI generate empathetic words, but it can also deliver them with a compassionate tone and cadence. (You can try it here… it is eye-opening to use a confused metaphor.)

For customer experience and marketing, that’s a big deal. It opens the door to truly empathetic automated interactions – think AI sales reps that sound enthusiastic and attentive or support chatbots that speak with genuine-sounding kindness.

But There’s Even More To It…

In the medium term, as consumers get used to more human-like AI interactions, we might see AI “personalities” representing brands – not just chatbots reading a script, but dynamic agents that adjust their tone based on the user’s mood.

Marketers and CX teams should start thinking about emotional design for AI: choosing the right empathetic style that aligns with their brand and audience. The bottom line: AI can now engage hearts, not just minds, so the best customer engagements may soon blend machine efficiency and human-like warmth.

AI as a Cybernetic Teammate

We often talk about AI as a tool, but it’s increasingly behaving like a team member. New research from Harvard and Wharton scholars (in collaboration with P&G) tested what happens when you embed AI into group work.

The findings were striking: individuals using an AI assistant (based on GPT-4) improved their performance so much that a single person with AI matched the output of a traditional two-person team without AI.

In the study, 776 Procter & Gamble professionals participated in product ideation workshops. Those working solo with AI support generated ideas that were on par with pairs of colleagues working together usually. Moreover, the best results came from two-person teams with AI—those AI-augmented teams were 40% more effective than individuals without AI and three times more likely to produce top-10% quality solutions.

Adding AI to the mix gave teams a vast creative boost and even helped balance out different expertise (sales folks and technical experts started producing more well-rounded ideas when both used AI). These outcomes earned AI the nickname “cybernetic teammate”—not replacing people but working alongside them, leveling them up.

Meanwhile, another report from the research non-profit METR highlighted how rapidly AI’s capacity for long-horizon tasks is expanding. They propose a metric for the longest task an AI can complete autonomously, and it’s growing exponentially – doubling roughly every 7 months. Extrapolate that curve, and in under five years, we could have AI agents that can, on their own, handle projects that take humans weeks.

Today’s best AI might struggle with multi-step projects longer than a few hours of human work. By 2030, an AI could feasibly manage a complex, week-long marketing project or an entire sales campaign execution. It sounds futuristic, but the trendline is clear.

Start rethinking workflows and team structure — NOW

It pays to treat AI as a capable junior colleague in the short term. For example, a copywriter+AI duo can produce as much content as two copywriters, so teams might pilot having smaller teams where each human is paired with an AI assistant.

In meetings or brainstorming, consider giving the AI “a seat at the table” to generate ideas in real time for the group to discuss. The P&G study shows this can elevate everyone’s output.

As AI systems handle longer tasks in the medium term, businesses might assign certain projects entirely to an AI (with human oversight checkpoints). We may see AI project managers who can draft a full marketing plan, schedule content releases, adjust based on analytics, and only ping humans for high-level approvals.

Professionals should prepare to supervise and collaborate with these autonomous agents. This could mean developing new skills – like prompt engineering, AI strategy vetting, and ethical oversight – akin to managing a talented yet unpredictable team member.

The upside is huge: teams augmented with AI can be far more productive and creative, as early research indicates. Companies that embrace AI teammates will likely outpace those that treat AI as just a fancy tool. The creative process is being “rewritten”—now, humans and AIs are co-creating, each playing to their strengths to achieve results neither could alone.

AI-Powered Search Shoots Through The Roof

A new report from Adobe Analytics indicates that traffic to U.S. retail websites from generative AI sources skyrocketed by 1,200% over the last twelve months. This unprecedented jump includes visits originating from AI-generated content, AI-driven product recommendations, and chat-based interfaces. The report also notes that generative AI integrations on social platforms—where users are served AI-curated shoppable posts—play a significant role in funneling traffic to retailers.

From a marketing standpoint, this data illustrates a major shift in how consumers discover products. Instead of solely relying on traditional search or direct brand engagement, shoppers increasingly use AI-driven content and recommendations as their primary entry point.

This new funnel can be particularly powerful when combined with personalized, AI-curated suggestions—allowing brands to catch potential customers at the earliest stage of their purchase journey.

AI brand visibility is critical

Yesterday’s game was SEO: making sure you are on the first page of Google. Tomorrow’s game is AI brand visibility: if Gemini, ChatGPT, or Perplexity doesn’t resurface your product, you are in deep trouble.

We recently published Brand Visibility in the Age of AI, a detailed guide on how to win the race for AI attention. Make sure you read it.

We are now working on another product that addresses this issue: how can you check and improve your brand visibility on AI? Stay tuned for that update.

Until next week — Peter and Torsten

Quick notes of the week

* Sponsored

• Perplexity started its AI Business Fellowship program last week. As we advance in the program (we are both members), we will share what we have learned with Horizon01 subscribers.

• We upgraded our Beehiiv account, allowing us to launch a much better web experience for our readers. Please look at our new site and let us know what you think.

• Speaking of Beehiiv, if you want to launch your newsletter, use our link and get a 30-day trial + 20% OFF for 3 months.*

• Want to take better notes with AI? Fireflies AI is a standalone AI notetaker for small teams and solopreneurs. An annual subscription gets you 40% off.*

• Our friend Jason Cormier is building the AI Marketing Forum, a free community of AI-enthusiast marketers. It’s a great place to meet with fellow AI-enabled growth professionals. You can join for free here.