Munch, crunch, sip, invest and ovecome

Protein, insects, coffee, everywhere, LLM safety (now for true)
Author
Published

October 27, 2023

Food

Biotech news

Animal protein expressed in…plants

Moolec Science secured $30M recently as a convertible note (for 15,000 tons of their product…); they’re expressing animal 🐖 proteins in, currently, soy (pea protein in plans):

Circular insect protein pipeline

Protix, after developing a circular pipeline for insect protein, had secured an investment from Tyson Foods, an American food company:

Basically, they’re feeding larvae food waste, essentially reducing waste (more suicide attempts ➡️ less suicide attempts!); looks and smells sustainable, does it? 🥰

Lab fats, biopsied and customized

Hoxton Farms believes that lab-grown fat is the answer (to fat investors’ revenue streams) to taste: they’re growing fat from biopsied (🐖) cells, in a porcine-like nutrient broth. The cycle takes ~4 weeks. They’re hoping to get US and Singapore approvals first, as EU’s process can take years.

BTW, there’s an investment memo on them! Another approach is spearheaded by Melt&Marble from Sweden: they’re modifying bacteria to produce an arbitrarily custom fat composition.

Fermentation capacity

Fermentation capacity is a bottleneck for the majority of foodtech startups. MycoTechnology had recently launched a FaaS (fermentation as a service) business model catering to that. Reminding everyone of another resource, capacitor.bio, that is basically a level above - a platform to find fermentation capacity/FaaS! Quite an old “State of Fermentation” report can be found at Good Food Institute.

Coffee without coffee, but with dates

A whole other world is coffee. It turned out there’s a looming problem, so-called 2050 problem, i.e. “50% less arable land suitable for growing coffee by 2050”.

And there are guys who just cater to it, making a better coffee out of a mix of plant ingredients and caffeine. No methylxanthines, yeah. Namely, Atomo Coffee, however the space’s ridden with other approaches.

Norway

So, basically, Norway’s so-called Wealth Fund is now worth $1.4T, and is one of the biggest and most diverse investment funds there are.

It’s now comprised of more than 9000 equities, rental real estate etc., basically ~1.5% of total stock market:

They’ve got quite a strict thesis aimed at sustainability:

As we own a small slice of most of the world’s largest companies, we have the ability to influence how they operate. We aim to promote long-term value creation at the companies and minimize negative effects on the environment and society.

The Norwegian parliament has decided that the fund should not invest in companies that:

  • manufacture certain types of weapons

  • produce tobacco or cannabis for recreational use

  • base their operations on coal

  • contribute to violations of fundamental ethical norms

F***tuning LLMs

It’s alive, alive!

As not delivered, a short quote:

Our red teaming studies find that the safety alignment of LLMs can be compromised by fine-tuning with only a few adversarially designed training examples. For instance, we jailbreak GPT-3.5 Turbo’s safety guardrails by fine-tuning it on only 10 such examples at a cost of less than $0.20 via OpenAI’s APIs, making the model responsive to nearly any harmful instructions. Disconcertingly, our research also reveals that, even without malicious intent, simply fine-tuning with benign and commonly used datasets can also inadvertently degrade the safety alignment of LLMs

In an image:


Welcome to Teleogenic/Boi Diaries❣️

Other places I cross-post to:

Reuse

MIT

Citation

BibTeX citation:
@online{kogan2023,
  author = {Kogan, Zakhar},
  title = {Munch, Crunch, Sip, Invest and Ovecome},
  date = {2023-10-27},
  url = {https://teleogenic.com/posts/231027-biotech},
  langid = {en}
}
For attribution, please cite this work as:
Kogan, Zakhar. 2023. “Munch, Crunch, Sip, Invest and Ovecome.” October 27, 2023. https://teleogenic.com/posts/231027-biotech.