Marco Fahmi

Mostly lurking.

[LSE Impact Blog @LSEImpactBlog]

✍️Dorothea Strecker, Heinz Pampel, Rouven Schabinger and Nina Leonie Weisweiler, explore how common data repository shutdowns are and suggest what can be done to ensure data preservation in the long-term.

#OpenData https://t.co/JJ7JB4DYHd

[Jeni Tennison @JeniT]

I am really troubled by the proposal from the Tony Blair Institute for a "National Data Trust" (NDT) for health data, particularly by the idea that the next government might actually go for it.

https://t.co/JEwy16hsuX

[giulio quaggiotto @gquaggiotto]

“When we have a technology that treats something as simple and fundamental as our name as an error, it robs us of our personhood" https://t.co/nIWg6tTv5O

[giulio quaggiotto @gquaggiotto]

Transformative governance of innovation ecosystems https://t.co/sGbOyE4tXH

[National Centre for Youth Substance Use Research @NCYSUR]

Congratulations team NCYSUR who is being awarded a 1.5 million dollar @nhmrc ideas grant, led by Dr Daniel Stjepanovic, @tianzesun , A/Prof Phong Thai, and Prof. @DavidHammondPhD.

[NHMRC CRE on Achieving the Tobacco Endgame @CREtobacco]

Congratulations 🎉 to CRE researcher Dr Carmen Lim @Cwernlim who has been awarded $660,000 by the @nhmrc to develop a program to understand how pro-vaping campaigns on social media influence young peoples' attitudes towards the use of e-cigarettes.

[Kevin Gee @kevg1412]

"He's not even a 10x engineer. He's like, 100x, or 1,000x engineer"

Gmail creator Paul Buchheit on how Bret Taylor once rewrote Google Maps in a single weekend:

[Axios @axios]

Behind the hype of generative AI, large companies are struggling to deploy the new technology — hitting cost and data management hurdles that are leaving many of their generative AI projects stuck in pilot phase. https://t.co/eUGASuBdCe

[Axios @axios]

As adoption of generative AI grows, providers are hoping that greater transparency about how they do and don't use customers' data will increase those clients' trust in the technology. https://t.co/XsUjlouhOi

[Rachel Woods @rachel_l_woods]

There's a resurgence of interest in fine tuning LLMs

I've yet to see a successful public use case where fine tuning > prompting.

But here's where I see fine tuning *mattering*:

First, fine tuning is for teaching an LLM specific tasks or behaviors

Not teaching an LLM new knowledge. For new knowledge, use Retrieval (store your data in an outside database and strategically pull the right chunks in to give the LLM context to your question)

But even in teaching LLMs specific tasks or behaviors - here's the catch...

LLMs are remarkably good at picking up tasks and behaviors from just a good prompt

THIS is what makes LLMs mind blowing after all

So that begs the question.

Where is fine tuning actually helpful?

Some use cases I could see developing are teaching LLMs tasks that are exceptionally difficult to describe, or fit into ~10 examples you can add to a prompt.

One way to think about this: if it would take someone a few weeks doing a task to 'master it' instead of being able to read training materials and get the picture...

That *may* be a use case for fine tuning

But proceed with caution

To truly teach an LLM a new behavior or task, you'll need to treat this like a machine learning project, not just throwing examples in and getting magic in return (which it still blows my mind that ChatGPT does this so well for us).

Things like:
- Dataset design
- Training and test data
- Overfitting
+ more as the tooling around fine tuning gets more sophisticated

The other obvious use case is cost.

If you can get a super small language model to do a task instead of GPT-4, there's meaningful cost savings there.

And if you're using a language model to do large scale tasks like triaging your customer support inbox, or analyzing public data for insights

The costs can add up.

But if you're wondering where the heck to invest in fine tuning...

My answer at the moment for most businesses is still:

Make sure you can't do it with prompts.

[Meredith Whittaker @mer__edith]

📢NEW PAPER!

Where @davidthewid, @sarahbmyers & I unpack what Open Source AI even is.

We find that the terms ‘open’ & ‘open source’ are often more marketing than technical descriptor, and that even the most 'open' systems don't alone democratize AI 1/

https://t.co/xRwEBc5LvO

[Bellingcat @bellingcat]

New updates on Bellingcat's #github this week. The 'whisperbox' API receives audio or video URLs and returns the video transcripts using OpenAI's Whisper model. Designed by Bellingcat discord member https://t.co/Q22L0Bi9Dv
Find the tool at: https://t.co/erjkSjtEfD

[Marco Fahmi @dataronin]

Opinion Paper: “So what if ChatGPT wrote it?” Multidisciplinary perspectives on opportunities, challenges and implications of generative conversational AI for research, practice and policy https://t.co/AcIVZgCcbN

[Marco Fahmi @dataronin]

Tackling generative AI’s sustainability problem https://t.co/v4CnGc2vpQ

[Daniel Severo @_dsevero]

Best poster award

[Global Investigative Journalism Network @gijn]

🗺️ NEW! Can journalists rely on #AI chatbots to successfully geolocate an image?

Investigative journalism group @bellingcat conducted 3️⃣ tests to examine the geolocation capabilities of #Bing and #Bard.

Reporter @DennisKovtun broke down the findings: ⬇️ https://t.co/4X8ca6xjVg

[Peter Griffin @petergnz]

Department of Internal Affairs has put out some guidance today on use of generative AI in the public sector...

[Marco Fahmi @dataronin]

What will the federal government do with generative AI? https://t.co/d4PP7N7M0D via @Nextgov

[Owen Boswarva @owenboswarva]

US judge finds flaws in artists' lawsuit against #AI companies https://t.co/aEwYL9fG1a #StableDiffusion #genAI #AIlaw #IPlaw #webscraping #openweb

Andersen et al v. Stability AI Ltd. et al https://t.co/zaGIUYYUi0 + https://t.co/I0OQrpP2aE

[Justin Alvey @justLV]

I “jailbroke” a Google Nest Mini so that you can run your own LLM’s, agents and voice models.

Here’s a demo using it to manage all my messages (with help from @onbeeper)

🔊 on, and wait for surprise guest!

I thought hard about how to best tackle this and why, see 🧵