The Rundown AI
Posts
ChatGPT catches holiday fever

ChatGPT catches holiday fever

PLUS: Microsoft's small but POWERFUL new model

Rowan Cheung
December 13, 2023

Sign Up | Advertise | Jobs | Tools

Welcome, AI enthusiasts.

It seems like ChatGPT is getting into the holiday spirit a bit early this year. Users are noticing the chatbot is becoming lazy — with data to support it.

According to the “Winter Break Hypothesis,” OpenAI’s assistant might’ve been trained to learn the human trends of laziness around the holidays. Let’s dig in…

In today’s AI rundown:

Is ChatGPT getting lazy near the holidays?
Microsoft ships small but powerful new model
9 new AI tools to try
How to use DALL-E 3 in any mobile app
Congressional candidate deploys AI for voter outreach
SI publisher fires CEO over AI controversy
Anthropic brings Claude to spreadsheets

Read time: 3 minutes

LATEST DEVELOPMENTS

CHATGPT

🎄 Is ChatGPT getting lazy near the holidays?

Image source: DALL-E 3

The Rundown: A new theory dubbed the "Winter Break Hypothesis" claims that ChatGPT has gotten lazier by learning that people relax more around the holidays — with statistically significant data to support it.

The details:

Complaints of ChatGPT laziness emerged in November, with users noticing shortened or denied responses.
OpenAI admitted awareness of issues, but no causality behind the change in model behavior.
The Winter Break Hypothesis suggests that since ChatGPT takes the date as input, trends around reduced human productivity this time of year impact its behavior.
X user Rob Lynch ran a test that fed GPT-4 prompts specifying different dates, finding much shorter responses in December.

Our thoughts: AI is just like us! Before we know it, ChatGPT will be throwing up an automated ‘Out of Office’ reply between Christmas and New Year’s. The subtle quirks of LLMs continue to amaze (and scare) many users.

TOGETHER WITH INCOGNI

🕵️‍♂️ Keep your info off the dark web

The Rundown: Every day, data brokers profit from selling your phone number, birthday and SSN (and you don’t want to know who’s buying it).

Incogni fixes this:

Scrub your personal data from the web
Confront data brokers head-on for you
Remove sensitive information from all broker types (even the scary ones)

Define your digital identity with Incogni. Click here to sign up and use code RUNDOWN60 for 60% off!

MICROSOFT

🤖 Microsoft ships small but powerful new model

Image source: Microsoft

The Rundown: Microsoft Research just released Phi-2, a surprisingly powerful 2.7B parameter ‘small language model’ capable of running on a laptop or phone — matching or exceeding the capabilities of larger models.

The details:

Phi-2 achieves impressive results among models under 13B parameters in reasoning, language, math, and coding benchmarks.
It outdoes the 7B parameter Mistral and Llama 2 models, and even surpasses Google's 3B parameter Gemini Nano 2 on some tests.
Microsoft credits strategic data selection emphasizing textbook-quality content, plus innovative knowledge transfer techniques in efficiently scaling up Phi-2.
But businesses can't tap the model yet, as Microsoft currently only licenses it for non-commercial research purposes.

Why it matters: Achieving these results in such a small package is quite the feat — and the ability to run on devices like a laptop or phone opens the door for more capable local, offline usage. With the release just a week after Google unveiled Gemini Nano, Microsoft is undoubtedly making a statement!

TRENDING TOOLS

🐻 BlitzBear- Outrank your SERP competitors in a few clicks. BlitzBear optimizes your blog SEO using AI without the manual work required (link)*

🔄 Recast Studio- Transform your podcasts effortlessly with AI (link)

💬 Chat with YouTube- Interact with AI to inquire about YouTube content (link)

💰 Mezzi- Minimize your tax expenses effortlessly (link)

📰 Brief- Get a custom AI-curated news digest (link)

📂 MagicDocs- Streamline your document management with AI (link)

🐦 Postwise- Craft viral Tweets in seconds (link)

🧠 AI-suggests- Generate tons of ready-to-use content and new ideas for your business instantly (link)

Browse the most popular tools ever featured with our tool database.

* = This is sponsored content

AI TRAINING

📱 How to use DALL-E 3 in any mobile app

Wish you had a way to harness the powers of AI image generation right in your favorite mobile apps? Now you can, with the help of the Microsoft Swiftkey keyboard (works on both iOS and Android).

Follow the steps below to integrate it on your phone by Paul Couvert:

Step 1: Download the Microsoft Swiftkey Keyboard (App Store, Google Play)

Step 2: Enable SwiftKey by toggling on the app under Settings → Keyboards

Step 3: Head to an app (iMessage, Twitter, etc.). Click on the globe icon to pull up your keyboards — you should now see a Bing icon. Tap create to access DALL-E and start generating directly in other apps!

In addition to DALL-E, you’ll also gain access to quick Bing AI chats, Microsoft Translator, an enhanced GIF selection and more. Happy AI image creating… now on the go!

QUICK HITS

📞 Congressional candidate deploys AI for voter outreach

Democratic congressional candidate Shamaine Daniels is deploying Ashley, the first AI phone banker in politics. The bot conducts personalized voter outreach using generative AI — with the ability to handle an infinite number of calls while using data on the candidate’s positions and key issues for voters.

QUESTION OF THE DAY

How do you feel about political candidates using AI for voter outreach?

Vote below (click for live results)

📊 Anthropic brings Claude to spreadsheets

Anthropic just announced that it’s Claude LLM is now accessible within Google Sheets via an extension and API access. The feature allows users to harness Claude’s writing skills via prompts directly weaved into a spreadsheet workflow. Access the new feature with Anthropic’s guide here.

🤯 SI publisher fires CEO over AI controversy

Sports Illustrated's publisher ‘The Arena Group’ just fired CEO Ross Levinsohn, following controversy surrounding the outlet’s AI-generated content. SI came under fire in November after a Futurism article exposed its use of AI authors, bios, and articles — scrubbing the evidence from the web post-inquiry.