😺 Claude 3 unleashed!

PLUS: this startup wants to DESTROY Google!

Welcome, humans.

There’s a fascinating debate on Twitter about whether AI should actually be considered as dangerous nukes, or if it’s essentially just “powerful math”.

Our take? We're crossing our fingers that AI isn't a civilization-ending threat like nuclear weapons (we're not keen on relocating to Los Alamos for coverage), and honestly, it's not that great at math either… :)

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

Anthropic released a significantly upgraded Claude 3.
Arc Browser is a new AI search engine with 100K+ daily users.
56% of Neuron readers want more AI in customer support.
ChatGPT can now read its output aloud to you.

Claude 3 is the first to match ChatGPT-4.

BREAKING NEWS: Anthropic just released Claude 3, and it’s the first chatbot as good as ChatGPT-4 (Gemini Ultra still lags both).

try Claude here

There are three versions of the model, and each is way smarter and more responsive than Claude 2:

Haiku is the fastest.
Sonnet powers the free Claude 3—this is *THE BEST* free AI chatbot available.
Opus powers Claude Pro, available at $20/month.

We played around with Claude 3 all of yesterday—here's our updated guidance on when to use it:

For formal/creative writing, use Claude 3. We literally write for a living—Claude 3 is now the best for crafting emails, drafts, memos, summaries, & more. It’s concise, nuanced, and human-like.

Claude 3 gives super creative responses for our entire bank of content creation prompts we use for testing.

ChatGPT-4 for data analysis. Although Claude 3 can handle a larger context limit of 200K (and 1M for select customers), it slipped up on numbers and didn’t design charts/graphs as well.

Claude 3 for coding (maybe). Some users report Claude 3 outperforms GPT-4 and specialized models like Grimoire in coding tasks, but detailed comparisons are TBD.

Additional insights:

Claude 3’s speed is standard, but slows down significantly during lengthy convos.
To build w/, Claude Opus is more expensive than GPT, but Sonnet and Haiku are cheaper (breakdown).
The knowledge cutoff is August 2023 so be wary of any questions relying on info since then (it’s not connected to the Internet).
You can compare all of these models on Poe.

Final thoughts:

We’re not saying ditch ChatGPT—instead, compare the tools to find which best suits your needs (we plan to use both).
OpenAI might be sitting on a big ChatGPT upgrade, and now would be a good time to release it.
AI models are being commoditized (i.e. there won’t be one clear winner).

FROM OUR PARTNERS

Never manually update your CRM again.

"I love updating our CRM,"...said no sales rep EVER!

At The Neuron, we've automated significant portions of our sales process using AI.

Here’s how you can, too:

Attention auto-updates your CRM based on sales calls and other customer interactions—say goodbye to manually entering MEDDIC, BANT, or any other data into your Salesforce, Hubspot, or Pipedrive…

This AI wizard can:

Log crucial data in your CRM.
Write follow-up emails.
Autoscore deals with AI.

Plus, it delivers insights that have pulled in an extra $250k in ARR per rep.

Start automating your CRM with Attention today!

This startup wants to take on Google.

Arc Browser is a buzzy new AI search app with >100,000 daily active users that wants to dethrone Google Search.

Two things we love after a quick test drive:

It searches the web and crafts custom webpages that summarize its findings—perfect for researching new topics.
Always-on blocker of ads, trackers, and banners.

Their launch video is a) masterfully produced, b) breaks down the product and why it matters:

Around the Horn.

ChatGPT can now read its responses—just tap “Read Aloud”.
Wix released a new text-to-website builder. Try it here.
*Acquire hundreds, even thousands of users overnight by sponsoring The Neuron.
India will mandate AI companies to get the govt’s permission before launching new models.
JPMorgan’s new AI cashflow management tool reduced manual labor 90% for some customers.

Tuesday Ticker.

Here are the results from last week’s poll:

B.P. selected Agree: “It beats having to listen to hold-the-line music for an hour.”
K. selected Disagree: “Not until it has been finessed more. Nothing worse than getting tied up with a Bot when they just keep going in circles. So frustrating!!”
B.N. selected Agree: “Less risk of wait-time delays, and potential language barrier challenges. Ideally, the AI can detect when a live-agent should be engaged based on the inputs and seamlessly hand-off. ”
K.C. selected Disagree: “Want to talk to a human. People are more empathetic and nuanced with interactions. Something a computer can’t do authentically.”

Here is this week’s poll: