Mihai Toader

Simple, Inquisitive fellow. Straight talker and curious about good life. Shipping is living.

[Mihai Toader @mtoader]

@cgst Maybe the baseline matters

[Sigma @sigmacomputing]

At Sigma, we recently closed our Series D round, securing an impressive $200 million. How did we do it? By leveraging the power of our own product to build an exceptional Data Room that wowed our investors.

#DataRoom #Fundraising #Investors

https://t.co/t3X0rgfrva

[Bogdan Gaza @hurrycane]

Introducing @datologyai — Making models better through better data, automatically!

https://t.co/LRKvyl9zRe

[Sigma @sigmacomputing]

Introducing Python in Sigma: #datascience for your business delivered in a single solution. Whether you write #Python or #SQL, work in spreadsheets, or code with #AI, Sigma gives diverse teams the autonomy to build together to deliver faster insights.
https://t.co/l5o5tVR67T

[Andrej Karpathy @karpathy]

# On the "hallucination problem"

I always struggle a bit with I'm asked about the "hallucination problem" in LLMs. Because, in some sense, hallucination is all LLMs do. They are dream machines.

We direct their dreams with prompts. The prompts start the dream, and based on the LLM's hazy recollection of its training documents, most of the time the result goes someplace useful.

It's only when the dreams go into deemed factually incorrect territory that we label it a "hallucination". It looks like a bug, but it's just the LLM doing what it always does.

At the other end of the extreme consider a search engine. It takes the prompt and just returns one of the most similar "training documents" it has in its database, verbatim. You could say that this search engine has a "creativity problem" - it will never respond with something new. An LLM is 100% dreaming and has the hallucination problem. A search engine is 0% dreaming and has the creativity problem.

All that said, I realize that what people *actually* mean is they don't want an LLM Assistant (a product like ChatGPT etc.) to hallucinate. An LLM Assistant is a lot more complex system than just the LLM itself, even if one is at the heart of it. There are many ways to mitigate hallcuinations in these systems - using Retrieval Augmented Generation (RAG) to more strongly anchor the dreams in real data through in-context learning is maybe the most common one. Disagreements between multiple samples, reflection, verification chains. Decoding uncertainty from activations. Tool use. All an active and very interesting areas of research.

TLDR I know I'm being super pedantic but the LLM has no "hallucination problem". Hallucination is not a bug, it is LLM's greatest feature. The LLM Assistant has a hallucination problem, and we should fix it.

</rant> Okay I feel much better now :)

[President Biden Archived @POTUS46Archive]

Today, APEC economies make up more than 60 percent of global GDP.

We are almost half of global trade.

The choices we make will matter to the entire world. So, it's up to us to harness the dynamism of our economies to realize a future that will benefit people everywhere.

[c0nc0rdance @c0nc0rdance]

In 2018, "Parachute use to prevent death and major trauma when jumping from aircraft: randomized controlled trial" was published in the British Medical Journal.

It sought to determine if wearing a parachute when jumping from a plane had any impact on survivability.

It did not.

[Massimo @Rainmaker1973]

Icelandic black sand is found on beaches near a volcano and consists of tiny fragments of basalt. This is the impressive effect created by strong winds moving the sand around, looking like "sand snakes"

[📹 Asa Steinars: https://t.co/2TOmN5Jv7K]

[Science girl @sciencegirl]

This baby skate at feeding time

[Science girl @sciencegirl]

New levels of chillin’

[steven monacelli @stevanzetti]

After posting surveillance footage of a self-driving Tesla causing a major pileup, it appears that Ken Klippenstein's account is being restricted from search results—except if you already follow him.

On the left is a photo from my account. On the right is a burner account.

[Sigma @sigmacomputing]

We’ve launched our Live Edit analytics collaboration feature! No more waiting on handoffs from others to *finally* make your changes. No more emailing data files. Work together in Sigma!
https://t.co/jvWBVOyK4l

[Mihai Toader @mtoader]

Voting matters. Or not. We shall see.

[Elon Musk @elonmusk]

Should I step down as head of Twitter? I will abide by the results of this poll.

[Mihai Toader @mtoader]

I'll probably end up breakin twitter's timeline service with so many blocks.

[Jeremiah Johnson 🌐 @JeremiahDJohns]

Elon Musk's inconsistencies, a thread:

[Mihai Toader @mtoader]

Before the Twitter Rapture i didn't have to block anyone (except one the one). Now I end up blocking several brigadiers daily. This will evolve well.

[Mihai Toader @mtoader]

We collectively beat the emergent AI (https://t.co/6BaHBb9Yc5 seems to time out a lot). That's how ants also win :)

[Mihai Toader @mtoader]

I think it's time to download my data.

[Massimo @Rainmaker1973]

Photographer Joel Sartore captured this footage of a Horsfield’s tarsier at Taman Safari in Indonesia. It's a carnivorous, nocturnal species of tarsier living in Borneo and Sumatra

[read more: https://t.co/sbuNAlBY4T]
[source: https://t.co/0z9jjR0wcD]

[Science girl @sciencegirl]

In the dry savannah and shrublands, a tiny bird, the Anthoscopus minutus builds its nest from spider silk and plant fibres

It builds a false entry and a blind chamber to avoid predation

Watch it close the ‘door’ as it leaves, leaving the false one ajar