😸 AI struggles with THIS

PLUS: Robotic iPad on wheels???

Welcome, humans.

Ugh, Monday again!? Luckily, we’ve got the perfect blend of corporate monotony and feudal fantasy to ease you back into the work week. Enjoy this AI video of everyone’s favorite workplace sitcom reimagined w/ an ancient Japanese twist:

If that’s not your thing, there’s also this montage of famous albums brought to life with AI… and this montage of famous memes brought to life by AI…

Happy procrastinating!!

Here’s what you need to know about AI today:

New study finds AI better at pattern recognition than rule application.
xAI's Grok-2 got a 2x speed boost after code rewrite.
Apple will roll out iPad-like home robot for 2026-2027.
Google AI Studio adds Prompt Gallery to test advanced AI features.

Advertise in The Neuron here.

New research reveals why you should always include examples in your prompts.

You know how most prompting advice tells you to include a few example outputs with your request? Well, a group of researchers from Amazon and UCLA just found out why that works as well as it does.

It turns out, large language models excel at learning patterns, even with only a few examples. What they don’t excel at? Applying rules—especially in unfamiliar scenarios.

Here’s what happened:

The paper looked at two types of reasoning in AI: “deductive” (applying rules) and “inductive” (figuring out patterns).
The researchers tested large language models like GPT-4 on a variety of tasks, from math to language to spatial reasoning.
They discovered that AI are actually really good at inductive reasoning, often getting perfect scores, but struggle with deductive reasoning, especially on “counterfactual” tasks (things they weren't trained on).

To test their theories, the researchers created a new method, called SolverLearner, which uses a two-step process:

Function proposal (the learning step): The AI is given a few examples and asked to figure out the underlying pattern or rule.
Function execution (the application step): Once the AI proposes a rule, it's tested using an external “verification” system (like a code interpreter), not the AI itself.

By using an external system to apply the rule, the researchers could focus solely on how well the AI learned the pattern without mixing in its ability to apply the rule.

This means future AI models (like Strawberry) will need to prove their deductive skills have specifically improved before they can claim anything close to AGI…

Here’s how to apply this in your own work with AI:

When asking an AI to solve a problem, play to its strengths and always give it a few examples.
Don't assume the AI can perfectly follow complex instructions or rules, especially with unique situations.
When possible, break down complex tasks into smaller pattern-recognition problems rather than long sets of rules.

In general, this paper is a good reminder to be aware that while an AI might understand a concept, it can still make mistakes when trying to apply it.

FROM OUR PARTNERS

Top sales teams still crush it in this economy. Attention is their AI secret.

Most sales reps are fed up with this economy (and aren’t you?). Too many deals are lost due to long sales cycles and budget cuts. Attention’s AI assistants can help.

Over 100+ sales teams from major companies like…

Snowflake
Datadog
Stripe

…all use Attention’s new AI Assistants to help close more deals and generate more revenue.

Check out one of our faves: Dax, a data analyst AI assistant, who:

Transforms call data into actionable insights
Identifies successful strategies for conversion
Creates performance visualizations for easy reporting.

If your business sells anything, we highly recommend checking Attention out here.

Around the Horn.

Meshy-4: Groundbreaking 3D Generative AI

Grok-2 got a 2x speed boost after 2 engineers at xAI rewrote the inference code stack from the ground up in 3 days.
VCs are eagerly buying into other VCs' AI startups at premium prices, sometimes 30% higher than recent valuations.
Apple is moving ahead with plans to build tabletop robots (rolling iPad that follows you around) sometime around 2026-2027.
There are now 70+ legit robotics companies hiring right now, signaling a boom-time for the industry.
Google has a new Prompt Gallery feature in its AI Studio to let you test out “long context, native multi-modal (image, video and audio), structured outputs, and more!”

Treats To Try.

*If you want to customize AI models on your own machine, don't miss Dell and NVIDIA's FREE webinar on September 12th. Save your seat here.
LM Studio released 0.3, an update to its desktop app to let you run local language models on your device, adding features like document chat, structured output API, and improved model management.
Paperguide combines research discovery, reading, writing, and reference management tools to streamline academic workflows for students and researchers.
Hexus streamlines the creation of product demos, videos, and guides by combining multiple content creation tools into a single platform.
Seven24.ai collects user feedback and converts it into actionable tasks, while also promoting positive reviews on external platforms.
My Journey visualizes your entire life on an interactive timeline, allowing you to explore memories, locations, and relationships through various views and filters.
SnapDiagram converts hand-drawn diagrams into editable digital versions.