5 min read

Weekend Briefing No. 28

Tech companies run out of money, Hugging Face tokens exposed, and how to install a Christmas tree
Weekend Briefing No. 28
Photo by Jamie Coupaud / Unsplash

Good Saturday morning and welcome to the last weekend briefing for 2023. I'm taking some time off to be with family and friends over the Holidays. I want to wish you and yours a Happy Holidays and a Happy New Year! See you in 2024.


Tech Start-Ups Run Out of Time and Money

It looks like the party is over for many startups. The money's run out and it's time to go home. I expect we'll see consolidation (i.e. mergers and acquisitions) in this space as large players gobble up some small players.

...approximately 3,200 private venture-backed U.S. companies have gone out of business this year, according to data compiled for The New York Times by PitchBook, which tracks start-ups. Those companies had raised $27.2 billion in venture funding. 

This "bust" reminds me a lot of the Dot-Com years when investors threw caution to the wind and invested in anything that had a ".com" in its name. Remember how that turned out? The moral of the story is that not every idea (crazy, sane, or otherwise) is a good one for a startup.


E.U. Agrees on Landmark Artificial Intelligence Rules

Coming on the heels of the Biden's Administration's AI rules, the EU finally agrees on their AI rules.

Policymakers agreed to what they called a “risk-based approach” to regulating A.I., where a defined set of applications face the most oversight and restrictions. Companies that make A.I. tools that pose the most potential harm to individuals and society, such as in hiring and education, would need to provide regulators with proof of risk assessments, breakdowns of what data was used to train the systems and assurances that the software did not cause harm like perpetuating racial biases. Human oversight would also be required in creating and deploying the systems.

I'm all for smart and sensible regulation that doesn't stifle innovation and protects privacy. It's been my observation that privacy tends to take a back seat to innovation in this space and these rules put those checks in place.

Some practices, such as the indiscriminate scraping of images from the internet to create a facial recognition database, would be banned outright.

META and IBM vs Google and Open AI

It looks like some of the titans in the AI industry are forming alliances to go after another group of AI titans. Meta (Facebook) and IBM are advocating an "open science" approach to AI.

Meta and IBM launch alliance advocating ‘open science’ approach to AI, putting them at odds with rivals like Google and OpenAI
The new AI alliance includes tech powerhouses such as Dell, Sony, chipmakers AMD and Intel, and several universities and AI startups.
Facebook parent Meta and IBM on Tuesday launched a new group called the AI Alliance that's advocating for an "open science" approach to AI development that puts them at odds with rivals Google, Microsoft and ChatGPT-maker OpenAI.

If you ask me, all of these titans have advocated for and created many open source projects. For example, Meta's PyTorch is used to train and fine-tune just about all LLMs out there. Google created Kubernetes, Go, and tons of other open source projects.

The rub here is the "open science" language. and I believe that is really about Generative AI. This is the opening salvo in the open vs closed LLMs and Generative AI wars and the players are building their teams.


Exposed Hugging Face API tokens offered full access to Meta's Llama 2

My partner works in Cybersecurity and based on what we talked about, securing your data and your proprietary information will be bigger than anything we innovate in AI. One successful attack and hack can easily compromise all your hard work.

Hugging Face API tokens exposed, major projects vulnerable
With more than 1,500 tokens exposed, research highlights importance of securing supply chains in AI and ML
Researchers at Lasso Security found more than 1,500 exposed API tokens on the open source data science and machine learning platform – which allowed them to gain access to 723 organizations' accounts.

Many AI models are vulnerable in ways that we never think of, for instance:

The researchers say that if attackers had exploited the exposed API tokens, it could have led to them swiping data, poisoning training data, or stealing models altogether, impacting more than 1 million users.

If you're a company that has a lot of technology and proprietary data, building a small attack surface will save you a lot of trouble later.


We all go wild for Bunco night

I wrote a short tale about dice probabilities, Bunco night, and stacking the odds.

A Tale of Dice Probabilities, Bunco Night, And Stacking The Odds
We all try to stack the odds in our favor.

Infographic

Every year a graphic similar to this one makes the rounds at Engineering firms. Granted, people celebrate the Holidays differently but I can't help but share these plans on how to properly install a Christmas Tree.

(c) Eng-Tips

Get on the list


End Notes

I wanted to give a shoutout to all my readers over 2023, you guys are amazing! Thank you! I couldn't have done it without you all! Although we fell short of the 1,000 subscriber goal, we saw amazing growth over the year. I wish you all a wonderful Holiday season and a joyous New Year! May 2024 rock for you!