simon_kun

Google launched similar functionality in NotebookLM today. You can generate podcasts from a wide range of sources: https://blog.google/technology/ai/notebooklm-audio-overviews...

Looks like you can generate from Website URLs if you add them as sources to your notebook, as well as Slides, Docs, PDFs etc. Anything NotebookLM supports.

possibly: request to be forgotten triggering model retraining if user data has been used to train it.

dma means permissions can be revoked iirc

I love blogs like this. reminds me of 2005.

Today.

This is an extremely insightful article. Strong agree with the core points made that these are copilots or co-processors rather than replacements. Maybe at some point in the future we will all be replaced (sound familiar?) but for now these are tools that can help us make progress as a society at a much faster rate

I did it through Dry January - it's socially normalized. Made it through Jan 2019, then just keep going.

doubtful - I'm tuning GPT3 with good midjourney prompts as we speak

Do folks know that 55% of ad revenue goes directly to the creators whose content you are watching? Apps like this sound great, but ultimately damage ecosystems. YT Premium is like $11.99/mo and you never see an ad again. If you think you're getting $12/mo USD value out of YT, then I question why you'd want to go through the hassle of this over just paying not to see ads (ps: premium income also goes to the creators!) - how much is your time worth?

this. if a Google/IG ads team was put on Twitter ads, their revenue and user satisfaction would likely go through the roof

How mission driven teams at FAANG build $1Tn+ market cap:

1. Product teams create utility at ~zero marginal cost per additional user

2. Platform teams drive distribution

3. Monetization teams convert utility to cash

4. GOTO 1

I'm by far not an expert in this space, but governance coins can also be utility coins - eg provide you with certain rights to use a service, or use a service at a discount or priority in some way ( eg. see Celsius). In this way, the coins circulate and can be used as a form of payment in an alternative to fiat but that were originally bought with fiat (similar to tokens at a laundromat). Overall, I recommend you look into DeFi for a few examples of this (eg Uniswap, Celsius)

I understand. Governance tokens are equity (usually with voting rights), and are used to generate a capital base for a decentralised product (think: crowdfunding). Not all governance tokens are airdropped and the secondary market for said tokens often results in them appreciating over time. I would suspect shapeshift to fund itself and generate wealth for the founders through doing this both at the point of issuance and over time.

they will release their own governance token through an airdrop. It's the new way to fund decentralised startups - no need for VC. Essentially it's the equivalent of an IPO on day 1.

'78 reporting in. We had a BBC Micro in our school library in Derbyshire. That coupled with the Speccys and c64's floating around at the time are probably a large part of why I'm sitting here in SF working at a senior level in big tech. Thanks BBC.

It's testament to Canonical that Ubuntu 8.04 LTS still gets security patches backported to 8.3. If you (still) have servers running Hardy, it's 'apt-get upgrade' time: http://www.ubuntu.com/usn/usn-1789-1/

Does anyone know of a patched version of rails 2.3 series on rubygems?

Hey Alastair, great post! Understanding OSM labels and so on can be pretty time consuming. You might like the mapdig library for CartoDB that lets you drill down into your spatial data to get a handle on what's what: http://vizzuality.github.com/mapdig/examples/osm_line.html (OSM london lines demo). Maybe next time :)

For those that are curious, the motherboard in these systems is a ASUSTeK P8H67-M PRO. Seems like it's a gaming board - understandable given it's an i7 platform.

I can't find one either. In fact, AWS doesn't seem to be running with ECC RAM except for some of the GPU instances. Happy to be corrected here.

I'm not sure AWS have updated their specs since then (all the specs for an ECU refer to a 1Ghz 2007 Xeon). We also tested our staging c1.xlarge and it scored less than the cloudharmony results indicated a c1.xlarge should.

If you mean tech as opposed to their data offerings, you should take a look at CartoDB http://www.cartodb.com. All the power of PostGIS with a lot of the pain points rounded off.

If you're interested in the relative performance of the current crop of JS engines, you may like to check http://arewefastyet.com/?a=b&view=regress for commit by commit benchmarking over time.

If people are interested in learning about vectors in a visual programming environment, then I highly recommend this tutorial on 2D vectors by Daniel Shiffman (of processing fame) http://processing.org/learning/pvector/

because when JSON is Gzip'd there's only around ~5% in it: http://goo.gl/b2Uur (the bit on geojson).