What's up? You're reading Inclined AI. A newsletter with more funk and soul than a Flea bass solo.

Here's the beat:

• ChatGPT is performing below its standard

• Google tests a tool to write news articles

• Semafor profiles the team at Tome

• Apple toys with the idea of their own AI

Be a sponsor of our newsletter.

SO, WHY DOES GPT-4 FEEL CLUNKY?

Let me tell you about the worst drift since Fast & Furious Tokyo Drift left theaters. It all relates to ChatGPT and the rumors about its drop in performance. 

A team of researchers from Stanford dropped a research paper this week that outlines startling findings. When they compared GPT-4 to its previous benchmarks in March, they found the model was drifting.

In other words, GPT-4 updates are hampering the model’s outputs.

It’s not a simple conundrum, though, because while GPT-4 did worse in math problems (dropping from 98% to 2% in performance metrics), GPT-3.5 improved.

The tricky riddle here revolves around how the teams develop these models. We don’t know what fine-tuning is going on behind the scenes.

The researchers found that GPT-4 fails to maintain a clear chain of thought when problem-solving. But did OpenAI do that on purpose to prevent costs from rising, or did something happen in a recent safety alignment?

Therein lies the rub, folks.

OpenAI keeps all this info and code to itself. Researchers don’t know what’s happening behind the scenes, which hampers GPT-4’s credibility.

Companies don’t want to run a model in their backend that will drift from top-notch to dumbo every other week. It’s not sustainable.

So keep an eye on OpenAI and see how they respond. Odds are they won’t even give this research paper a second thought.

Someone else must replicate the results before they do anything about this.

You’re already learning all you can about AI; why not take it a step further?

I’ve been reading this newsletter, The Strategy Deck, and find that it’s chock-full of market insights and competitive analysis that keeps me keen on everything AI.

Writer Alex Sandu has over a decade of experience with tech market research and product management and fills every edition with brilliant business and product strategy insights. Yes, that’s right, it’s not just another AI newsletter.

Find out what AI platforms will give you a competitive advantage and find relevant, actionable information in every edition.

Subscribe for free!

For dedicated market research and product strategy support for your AI application, reach out directly to Alex - here or alex @ thestrategydeck [dot] com

GOOGLE IS COMING FOR MY JOB

Here’s a secret: The Associated Press uses AI to write corporate earning reports for publication and has been for years. AI in the newsroom is not new, but Google wants to up the ante.

Anonymous sources told The New York Times that Google pitched their execs and other big newsrooms on a new project called Genesis.

Genesis is generative AI made for journalists.

Google wants it to be a helpmate and not a replacement. Think Clippy, but for breaking news, that’s where they’re going with this.

Here’s the thing, I think it’s a good idea, and I think you should, too.

Google pioneered the transformer model and can align AI to specific niches. If Genesis helps journalists maintain accurate reporting and write better stories, newsrooms should adopt the practice.

AI like that can help local publishers fit more stories into a local print and serve their community with broader coverage. A journalist with less on their plate is advantageous.

Genesis might be the solution to helping reporters bring attention back to their articles.

But that’s if the tool is a supplement, not a replacement.

If they want to take humans out of the equation, it won’t work. There’s too much fact-finding and detailed writing to hand everything to a bot. 

However, can you imagine an investigative reporter with an AI helper spotting trends they might miss? We’d uncover more and get better news from our favorite outlets.

You can’t build the entire boat out of AI, and Google knows that.

Quick Nuggets

🍎 Apple starts running an internal AI off their model, Ajax

💄 Maybelline's new beauty app partners with Microsoft

🤝 An armistice in the AI wars could be around the corner

📚 Don’t worry AI won’t disrupt books as much as you think

🗣️ Langauge learning with AI is better when you use these tips

🤔 Can we trust any review we read now that AI is this popular

🙋 Helsing AI is working in warfare but promises they are the good guys

⭐️ A quick feature on the founders of Tome and their mission

Check Out Our Sunday Edition

You can dive into more AI news and topics with us every week by subscribing to our premium edition.

We’ve written about the following:

• Here Is The Work AI Will Replace Soon

• 8 Role Prompts that Still Work on ChatGPT

• 10 AI Tools You’ll Actually Use

If you’re not already subscribed, that’s okay. We’re offering a free 7-day trial so that you can read this one. That’s how excited I am to post it.

Claim the offer below to support our writing 🙂

🔥 Fresh Products

• Spoke - powerful & secure AI summarization (link)

• Swiftbrief - SEO briefs in 2 mins w/ accurate keyword data (link)

• LangSmith - build & deploy LLM applications w/ confidence (link)

• Second Nature - employee training that is enjoyable (link)

• MailerLite - AI-assisted emails built in seconds (link)

• Belva - AI phone calls for everyone (link)

• Arvin - 24-hour assistant, standby anytime on any website (link)

• BeeBee - AI analyzes public companies (link)

• Unsummary - summaries of books, movies, tv-shows, podcasts (link)

Good Content, Cheese Car

Not sure how I should feel about this. I think I should be scared? I do have a dairy allergy, after all…

_{- That’s it for today. I hope you enjoyed the latest edition of inclined.ai - Davis.}