- AI Tool Report
- Posts
- š¦ Meta launches biggest AI model ever
š¦ Meta launches biggest AI model ever
šØ Mastercard in AI pricing scandal
TOGETHER WITH LATELY
Welcome to AI Tool Report!
Wednesdayās top story: Meta has launched its newest AI modelāLlama 3.1āwhich is the biggest AI model, to date.
š¤ļø This Morning on AI Tool Report
š¦ Meta launches biggest AI model ever
šÆ How to transform your social media advertising with Lately
šØ Mastercard in AI pricing scandal
ā How to automate your task management
š World-first: Autonomous tandem-drift cars
š³ļø Should Tesla invest $5B in AI?
Read Time: 5 minutes
STOCK MARKETS
![Stock tracker](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/a09c311e-aa19-4708-adf0-ce4d3ebec865/430.16__37___1_.png?t=1721804529)
š AI stocks remain unchanged during Tuesdayās session. Microsoft seems to have navigated the tech failure without damaging the stock value. Since the news broke last Friday, Microsoft has actually gained 2.79%. Learn more.
ā ā ā ā ā ā
LAUNCHES
![Meta launches biggest AI model ever](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/2a31c751-a765-490e-8ea2-aca2c1e352b3/Meta_launches_biggest_AI_model_ever__1_.jpg?t=1721813785)
Our Report: After teasing that it was working on an open-source AI model, that would outperform leading AI modelsāfrom favorites like OpenAI, Anthropic, and GoogleāMeta has released Llama 3.1, which is the largest open-source model in the industry, to date, and can be downloaded by anyone (with a big enough infrastructure) from Hugging Face, GitHub or Meta itself.
š Key Points:
Llama 3.1 isnāt multimodal and just handles text (Meta is working on this), but it can perform coding tasks, answer math questions, and summarize documents (including PDFs/spreadsheets) in 8 languages.
Llama 3.1 was trained and developed on new techniques and outperforms OpenAIās GPT-4o at executing code and generating plots, but early tests show itās weaker in multilingual capabilities and general reasoning.
Itās been trained on the same dataset as previous modelsānon-English, code, Math, web, and synthetic data (created by other AI models)ābut Meta has applied ārigorousā quality control and filtering to improve accuracy.
š¤ Why you should care: Meta hasnāt confirmed exactly where it got its training data from (outside of public web pages and web files), claiming āit's the secret sauce that goes into building these models,ā giving it a competitive advantageāwhich is the approach most other tech giants take tooābut critics believe that the secrecy is actually a tactic to delay the inevitable copyright lawsuits that are coming, as it notoriously trains its AI on Instagram and Facebook posts, photos and captions, making it difficult for users to opt-out.
Why is Meta not disclosing where it got its data from? |
ā ā ā ā ā ā
Together with Lately
Lately is the first Deep Social Platform that merges groundbreaking neuroscience with cutting-edge AI to transform your social advertising.
Latelyās Neuroscience-Driven AIā¢ learns and adapts to your unique brand voice to create social content your audience will love.
Repurposes long-form content into high-performing social posts
Elevates your content from meh to extraordinary
Creates a customized voice model to reach any target audience
Populates your social media calendar for months on end
All faster than you can grab a cup of coffee ā
TRENDING TOOLS
Guidde is the secret presentation tool that will 10x your teamās productivity š¤«š”
Meco is a distraction-free space for reading and discovering newsletters, separate from the inbox
Syncly surfaces customer pain points from everyday communications
Protide enhances designs with real-time AI-driven iterations
Qolaba unifies multiple AI models from industry leaders
ā ā ā ā ā ā
PRICING
![Mastercard in AI pricing scandal](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/f8e82674-d3e8-4dcd-bb24-30b283d1c5a3/Mastercard_in_AI_pricing_scandal.png?t=1721814196)
Our Report: The Federal Trade Commission (FTC) is investigating 8 companiesāincluding Mastercard, Revionics, JPMorgan Chase, Accenture, and McKinsey & Coāabout how theyāre using AI to change pricing, based on customer behavior and characteristic data, a practice known as āsurveillance pricing.ā
š Key Points:
The FTC wants information on how they collect customer data, who their customers are, how their customers use their products or services, and how their pricing strategies are affecting consumer pricing.
They believe that āAmericans deserve to know whether businesses are using consumer data to deploy surveillance pricingā, and are exposing practices that could āreshape how consumers purchase goods and services.ā
Mastercard has agreed to ācooperate with the FTC ā and Revionics has denied that it uses AI to target prices at individuals and ādoes not conduct operations related to the surveillance of consumers.ā
š¤ Why you should care: Although several companies are experimenting with pricing models that change prices based on different variables (eg. JetBlue applies āpeakā and āoff-peakā pricing to its checked-in bags and Amazon has an algorithm that gauges how much it can raise its prices before its competitors raise thereās), surveillance pricing is a āshadowy ecosystem of pricing middlemenā because it uses personal, consumer information to determine how much people are willing to pay for a product.
AUTOMATION OF THE WEEK
ā ā ā ā ā ā
Managing tasks efficiently is crucial for productivity. This weekās automation turns Slack messages into Trello cards, streamlining projects and task management, ensuring nothing falls through the cracks.
Want to know how to set this up in detail? Read this
BuzzFeed, after receiving many, many applications, teamed up with Uncubed to implement IBM Watson Candidate Assistant. This AI tool engages applicants in tailored career discussions, suggesting suitable roles. Trained on BuzzFeed-specific queries, it provides real-time answers and links to open positions.
The outcome is remarkable: 87% of AI-assisted candidates progress to interviews, a 64% improvement. This AI-driven approach helps BuzzFeed focus on quality candidates and optimize their recruiting strategy.
BREAKING NEWS
ā ā ā ā ā ā
AUTOMOBILE
š World-first: Autonomous tandem-drift cars
The Toyota Research Institute and Stanford Engineering have used AI to build the worldās first autonomous tandem drift cars that can ādriftā, in harmony, alongside each other, following each other's lines.
The first car in the drifting tandem uses a precise amount of throttle, brake, and steering, which the second can read and automatically respond toāthe AI can solve/resolve an issue 50x a secondāpreventing it from crashing.
The physics of ādriftingā are similar to how a car behaves on snow and ice and this technologyāwhich controls the car like an F1 driverācould be used to manage a loss of control when driving on slippery roads.
ā ā ā ā ā ā
DEALS & PARTNERSHIPS
š³ļø Should Tesla invest $5B in AI?
After reporting Teslaās 4th quarter of disappointing profit, Elon Musk posted a poll on X, asking his followers whether Tesla should invest $5B into his AI company, xAI, to ātest the waters.ā
This also comes after Musk previously expressed plans to explore a partnership between xAI and Tesla and believed Tesla should be seen as an AI or robotics company rather than just an auto company.
With 23hrs remaining, the majority of voters are in favor of the proposed deal, but Musk has stated heāll need shareholder approval and the valuation of the deal would have to be āset by several credible outside investors.ā
RECOMMENDED RESOURCES
We read your emails, comments, and poll replies daily.
Hit reply and tell us what you want more of!
Until next time, Martin & Liam.
P.S. Donāt forget, you can unsubscribe if you donāt want us to land in your inbox anymore.
What did you think of this edition? |