😺 Lazy ChatGPT

PLUS: China just entered the chat.

Welcome, humans.

We touched on this briefly yesterday, but it's official: The Neuron has surpassed the 250,000 subscriber mark! It's a significant milestone for us, yet, echoing the legendary Kobe Bryant, "Job's not finished."

P.S. We're on the hunt to acquire other AI newsletters. If you're considering selling, slide into our DMs!

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

  • OpenAI is giving ChatGPT-4 Turbo a boost to shake off its laziness.

  • ChatGPT-3.5 got an upgrade and price cut, too.

  • There’s a new open-source model topping leaderboards.

  • AI-generated explicit images of Taylor Swift surfaced on X.

OpenAI: we’re making ChatGPT-4 less lazy.

Almost every single ChatGPT power user we talk to has the same gripe: ChatGPT-4 has become way lazier and dumber over the past 6 months.

We agree.

It’s like that exceptionally bright kid who just wouldn’t answer questions in class (you know…that person you probably work for now)…

Last month, OpenAI pledged to address this issue, and yesterday they announced an upgraded GPT-4 Turbo model specifically “intended to reduce cases of ‘laziness’ where the model doesn’t complete a task”.

TBD if this truly amps up performance—in the interim, we recommend using the old version of ChatGPT a la ChatGPT Classic GPT OR trying some of these prompting hacks to boost response rates.

In other news…OpenAI announced a smorgasbord of updates for developers. Notably, GPT-3.5 Turbo (the free ChatGPT version) is getting an upgrade and significant price cuts for input/output tokens (50% and 25%, respectively).

They also rolled out two new embedding models and a moderation model aimed at helping developers identify harmful content.

Pirates, hoist the sails – it's prime shipping time!

FROM OUR PARTNERS

Your business should be talking more about Speech Intelligence. 

Think about it: speech is our most natural way of communicating—we’ve been doing it for 100,000 years, way before ChatGPT came along. 

So, why are we still stuck pounding away at keyboards like some sort of primitive species?

Speechmatics brings the latest breakthroughs in Speech Intelligence to your business.

Combining AI and LLMs, Speechmatics is your go-to for transcribing, translating, summarizing, interpreting and analyzing the spoken word, in 40+ languages.

All in real-time and loved by brands like Deloitte, Red Bee Media and Nvidia.

Forget the clunky "Hey Siri" interactions; Get started with Speechmatics for free today. 

🚨New AI model alert!🚨

Meet Yi, the brainchild of Chinese AI startup 01.AI (which just hit unicorn status).

According to benchmarks, Yi is the best *multimodal, open-source chatbot* on the market. In other words, Yi can interact with text, voice, & images AND its code is publicly available.

01.AI

It’s also the best available model (that we know of) to come out of China, which has significantly trailed the US in AI thus far.

Yi-34B shows particular promise in coding queries, but we're on standby for more juicy details.

Hey, so sorry to interrupt…

…but we’ve got an offer for you: refer one friend to The Neuron and you might win $250. 

more referrals = better chance of winning.

TBH, not many people have done this yet, so the competition’s lighter than you'd think. GO GO GO:

Around the Horn.

  • AI-generated sexually explicit images of Taylor Swift went viral on X, with some posts getting 24K retweets before being removed.

  • The Ray-Ban Meta smart glasses’ new AI powers are impressive, and worrying (link).

  • HuggingFace will offer access to its open-source models through Google Cloud in a new partnership.

  • This research paper overviews all of the open/close LLMs on the market, including GPT-4 and Llama.

Intelligent Insights.

We might be the wittiest writers when it comes to AI, but there are definitely some brains who out-IQ us.

  • I literally spoke with Nvidia’s AI-powered video game NPCs (link).

  • You Can’t Truly Be Friends With an AI (link).

  • How Not to Be Stupid About AI, With Yann LeCun (link).

  • Audio deepfakes emerge as weapon of choice in election disinformation (link).

  • Don’t Fuss About Training AIs. Train Our Kids (link).

Friday Faculty.

  1. 7 technical/non-technical positions at Deci (New York/Tel Aviv/Remote).

  2. Machine Learning Intern (MS/PhD) at Tiktok (San Jose).

  3. HW/SW Co-Design Engineer at OpenAI (San Francisco).

  4. 78 open jobs at Scale (all over the place). 

  5. 15 roles at Adept AI (Bay Area).

Search from over 10,000 AI roles to find your perfect fit on The Neuron’s Jobs Board.

A Cat's Commentary.

That’s all for today, for more AI treats, check out our website.

Get your brand in front of 250,000+ professionals here.

See you cool cats on Twitter: @nonmayorpete & @noahedelman02

What'd you think of today's email?

Login or Subscribe to participate in polls.