- Horizon AI
- Posts
- 'Do not trust OpenAI', warns Edward Snowden 🤯
Welcome to another edition of Horizon AI,
US whistleblower Edward Snowden has warned against trusting OpenAI following a recent appointment.
Let's jump into it!
Read Time: 3.0 min
Here's what's new today in the Horizon AI
Snowden Sounds Alarm on OpenAI after Former NSA Chief Joins Board
Nvidia Launches Open Model That Creates LLM Training Data
AI Tools to check out and AI Job board
The Latest in AI and Tech 💡
AI Showcase
📣 Help Shape Our Content!
We're creating more content about AI and automation. Please vote in our poll today to let us know which topics you find most interesting and what you want to learn more about.
Your feedback helps us make better content for you!
What type of content would you like to see the most? |
AI News
AI SAFETY
Snowden Sounds Alarm on OpenAI after Former NSA Chief Joins Board
Edward Snowden, the US whistleblower who leaked classified documents about surveillance in 2013, has urged the public not to trust OpenAI or its products following the company's recent appointment of former NSA director Paul Nakasone to its board of directors.
Details:
Nakasone has joined OpenAI's newly formed Safety and Security Committee with the role of helping to utilize AI for cybersecurity purposes.
Snowden calls Nakasone's appointment a "willful, calculated betrayal of the rights of every person on Earth," suggesting OpenAI intends to leverage its AI for mass surveillance.
He believes the combination of advanced AI capabilities with the vast amounts of surveillance data accumulated over the past two decades will "put truly terrible powers in the hands of an unaccountable few."
Finally, he criticized the "AI safety panic" over a threatening superintelligence, arguing that a greater danger lies in incorporating political views into AI models to generalize minority opinions, citing Google's recent controversy over diversity rules in its generative image AI, which led to historically inaccurate depictions.
NVIDIA
Nvidia Launches Open Model That Creates LLM Training Data
Source: Nvidia
Nvidia has released its Nemotron-4 340B, a family of open models designed to generate high-quality synthetic data for training and fine-tuning large language models. The company claims its instruction model can match or even outperform GPT-4 in certain benchmarks.
Details:
The Nemotron-4 340B family consists of a 9 trillion token base model, an instruction model for generating diverse synthetic training data, and a reward model to filter high-quality AI outputs.
Nvidia claims synthetic data can improve data quality and augment limited real-world datasets, benefiting commercial LLM applications across industries like healthcare, finance, and retail.
In benchmarks such as MT-Bench, MMLU, and HumanEval, Nvidia's instruction model generally outperforms other open-source LLMs like Llama and Qwen, and in some tests even surpasses GPT-4's performance.
Nvidia is making the models available under an open license, allowing commercial use and customization of the language models.
Nvidia isn't aiming to make Nemotron compete with Llama 3 or GPT-4. Instead, they designed the model family to assist developers in training better or more models across various domains.
AI Tools to check out
🦾 Teamcreate: Build your dream team in minutes.
🚀 Sixfold: Generative AI solution designed to boost underwriting capacity, accuracy, and transparency for insurers, MGAs, and reinsurers.
📧 Leap's Lead Research: Enter an email to find out everything you can about that person.
⚡ Simplescraper: Extract data from any website in seconds.
✨ Everlearn’s mindmap generator: Create, organize, and visualize your ideas effortlessly with this mindmap generator tool.
AI Job Board
💻 Product Data Science - Shopify
📈 Backend Engineer - You.com
👨💻 Product Manager - Hopper
💻 Software Engineer - Airbnb
The latest in AI and Tech
Meta has halted the rollout of its AI assistant in Europe after Ireland's privacy watchdog raised objections to the company's plans to train its language models using public data from Facebook and Instagram users. The tech giant claims regulators' restrictions could result in an inferior product for European users compared to other regions, citing the need for "local information" to deliver a quality experience.
McDonald's is ending its two-year AI experiment for drive-through orders after piloting the system at over 100 locations. The fast-food giant, which launched the automated order taking (AOT) in partnership with IBM, will shut it down on July 26, 2024. McDonald's still sees potential in this technology and plans to partner with a new company by year-end for further research.
While Apple Intelligence features won’t be out until 2025, Apple is planning to upgrade Siri with improvements to natural language processing and other key features before the end of 2024.
Camb AI has launched Mars5, an open-source voice cloning AI model that promises higher realism compared to competitors like ElevenLabs. The company claims Mars5 can capture nuances in speech such as emotion, rhythm, and intonation. Camb is also developing Boli, a translation model aimed at better handling context and colloquialisms than tools like Google Translate.
AI Showcase
🖼 AI art of the day
Source: ashish2410 on Midjourney
A high-resolution cartoon style illustration of Small-town Old railway station with a single train arriving, the platform lit by vintage lamps and a cool breeze blowing, blue sky, white clouds, afternoon, rendered in a colorful cartoon style, with bright colors and clear lines, well detailed, colorful, eye-pleasing --ar 4:5 --stylize 500📹 AI video of the day
⚡Verdant Ruins Quest⚡
Generated with @LumaLabsAI Dream Machine
Workflow below:
— Heather Cooper (@HBCoop_)
10:03 PM • Jun 16, 2024
That’s a wrap!
We'd love to hear your thoughts on today's email!Your feedback helps us improve our content |
Not subscribed yet? Sign up here and send it to a colleague or friend!
See you in our next edition!
Gina 👩🏻💻