😺 AI is not that good (yet)

PLUS: What the future of AI will look like...

Welcome, humans.

These days, it feels like everyone’s claiming their products and services are “AI-powered”. Even coffee roasters (seriously).

It's like slapping a "gourmet" label on a can of Spam to make it seem like a high-end culinary choice. Needs to stop!

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

Sam Altman remarks that today’s AI is “not that good”.
Big Tech is lowering its expectations for AI-driven growth.
Microsoft, Google, and OpenAI are developing AI agents.
OpenAI filed to dismiss Elon Musk’s lawsuit.

AI is decent, but it’s not about to add an extra “0” to your profits.

Despite the hyperbolic forecasts we often hear from tech CEOs about AI during earnings calls, here’s a reality check: AI in 2024 isn’t all that impressive.

Yeah, we said it, don’t @ us.

That’s not to say it’s worthless—if leveraged properly, today's AI can boost productivity by 20-40% in areas like coding, writing, and HR.

Quick plug: Take our Intro to ChatGPT Course if you’re not yet seeing these productivity boosts.

But today’s AI isn’t up to snuff for replacing most jobs: it struggles with complex tasks, produces errors, and can’t even unmute itself on Zoom (like your folks).

The reason we’re pointing this out is that many folks (read: Nvidia investors) are dreaming too big about how soon AI will truly revolutionize things. So much so, that even Big Tech is scaling back its own hype:

“How long can Wall Street’s artificial intelligence-fuelled rally continue without clear evidence that generative AI is giving a meaningful lift to business?”

Richard Waters, FT

Amazon CEO Andy Jassy told investors in February that near-term AI revenue is “relatively small.”
Salesforce execs said its AI offerings wouldn’t make a “material contribution” to revenue growth in 2024.
Microsoft reported that its Azure AI platform's revenue made up only ~1% of the company’s expected total revenue.

Even Sam Altman, the guy building the damn thing, admits AI is a work in progress:

"I think right now expectations are extremely high. Reality is still pretty bad. Honestly these models are not that good. I think very quickly expectations will start to come down…”

Sam Altman on 20VC

Here’s our take: Obviously, this is not where AI stops. It’s just important to be honest with where we’re at. The models will continue to get “really really good”.

And beyond chatbots, we’re convinced that what will truly transform business and unlock trillions in productivity are AI agents (more on this later)...

FROM OUR PARTNERS

How to choose the right LLM for your business.

You get the point: AI can help supercharge your organization and boost productivity yada yada...

There’s one problem: which model should you use? Claude? Jurassic? Cohere?

Learn how to select the right LLM by signing up for AWS and Arthur AI’s webinar “Not all models are created equal: Key factors to consider when selecting an LLM.”

They’ll show you exactly how to evaluate models based on key factors like speed, accuracy, transparency, and, yes, cost.

AI builders shift their focus to agents to generate material $$$.

say agents without saying agents

OpenAI and Google, AI’s two frontrunners, similarly believe that chatbots aren’t where AI ends—they’re both zealously working towards AI agents that companies will shell out big bucks for.

WTF is an AI agent?!

From March: “If today’s chatbots are like rearview cameras on a car, agents are like automatic parallel parking.” In other words, AI agents can do multiple actions of a job by itself.

For instance, Microsoft is planning an AI agent that can…

→ spots when a large order hasn't been processed by a customer.
→ drafts an invoice for that order.
→ ask the business whether it wants to send the invoice.
→ follows up on the customer’s response and payment.
→ records everything in the company’s database.

Similarly, OpenAI is building an AI agent that can manage your desktop applications, taking on tasks like completing expense forms and updating accounting records, or shifting data from a document to a GSheet.

Why it matters: AI agents represent the next big leap in workplace automation—systems that don't just assist but fully automate job roles, slashing task completion time from days to mere minutes.

Of course, we’re not there yet. Today’s AI agents, like Devin, are promising yet not complete. The transformative '10x' agents might not arrive until 2026 or later. It’s anyone’s best guess.

Up next? ChatGPT-4.5. Or 5. We don’t expect these models to constitute agents, but we won’t be surprised if they completely outshine everything we’ve seen so far.

Around the Horn.

Llama 3 70B narrowly matches Claude Sonnet and Gemini Pro but falls short of ChatGPT-4 Turbo and Claude 3 Opus in the LMSYS Chatbot Arena Leaderboard.
A new study found that ChatGPT-4 surpasses human doctors in medical board residency exams across various specialties.
Apple is reportedly preparing an AI that operates directly on its devices, not via the cloud.
OpenAI filed a motion to dismiss Elon Musk’s lawsuit.

Treats To Try.

Sonnet

*Bland.ai lets businesses deploy AI phone agents capable of taking 1,000,000+ phone calls at once. Try calling the AI here! Or sign up here for something crazier(!)
Grimo is a new kind of AI-powered notebook that integrates content from YouTube, podcasts, and more all in one spot.
Sonnet is an AI meeting assistant that preps you for calls, summarizes discussions, and updates your CRM post-meeting.
The Pipe is an open-source API that leverages GPT-4V to help you decipher complex documents—be it a PDF, Word doc, webpage, or any image codebase.
SkillexExchange is an AI-powered job board that uses smart filters to help you find your ideal job.

*This is sponsored content. Advertise in The Neuron here.