What about DiskANN? https://milvus.io/docs/disk_index.md
221 karma · joined February 5, 2012
Summary:
* Spotify’s Layoff Impact: CEO Daniel Ek was surprised by the significant operational challenges following the layoff of 1,500 employees.
* Financial Performance: Despite achieving record profits of €168 million and revenue growth, Spotify missed its targets for profitability and user growth.
* Investor Reaction: Shares increased by over 8% after the earnings report, reflecting investor confidence.
* Long-Term Strategy: Despite short-term issues, Spotify believes the layoffs were necessary for long-term profitability.
I don't think the answer is one or the other at all like the blog post is implying. I think his previous conclusion of a threshold makes more sense. Perhaps though they have chosen the wrong abstractions and hence the confusion
1. Lucene is trying to get Approximate Nearest Neighbours (ANN) search working for semantic search purposes: https://issues.apache.org/jira/browse/LUCENE-9004 https://github.com/apache/lucene/issues/10047
2. The Panama Vector API allows CPU's that support it to accelerate vector operations: https://openjdk.org/jeps/438
So this allows fast ANN on Lucene for semantic search!
How did people do this before Lucene supported it? Only through entirely different tools?
On first look this is genius but it seems pretty tautological in a way. How do we know if the explainer is good?... Kinda leads to thinking about who watches the watchers...
Edit: The answer was patience
How many finance workers on wall street do you think provide direct value to society?
Thank you!
I feel like it's missing some things though:
* Data Entry -- OCR and the like
* Data Retrieval -- Don't search engines still qualify as AI
* Sorting mail?
* Other factory use cases like removing undesirable tomatoes: https://www.youtube.com/watch?v=aYQ_5c6m8Is
* Many others I'm not thinking of...
Also, is there a way to do this where they won't be able to patch? Maybe alternate domain names for these known domain names that are being blocked, similar to what thePirateBay did.
1. Rate of progress on AR/VR at meta
2. People within Meta not "Giving a Dam!" / Poor decisions made within the org
3. Poor quality of the execution which I guess is related to #2
Relevant quote: "It pains me to hear people say that they don't even get their headset out to show off at the company because they know it's going to be a mess of charging and updating before they can make it do something cool," Carmack said at the time. "VR should be a delight to demo for your friends."
"- The two important but different abilities of GPT-3.5 are *knowledge* and *reasoning*. Generally, it would be ideal if we could *offload the knowledge part to the outside retrieval system and let the language model only focus on reasoning.* This is because: - The model’s internal knowledge is always cut off at a certain time. The model always needs up-to-date knowledge to answer up-to-date questions. - Recall we have discussed that is 175B parameter is heavily used for storing knowledge. If we could offload knowledge to be outside the model, then the model parameter might be significantly reduced such that eventually, it can run on a cellphone (call this crazy here, but ChatGPT is already science fiction enough, who knows what the future will be)."
& "Yet there was a WebGPT paper published in Dec 2021. It is likely that this is already tested internally within OpenAI."
It definitely feels like this may be the next step in making this kind of system robust. It ends up being an interface for search.
So many data scientists are full of themselves thinking they are magicians and software developers are blacksmiths who are beneath them.
Incrementally at my company the SWE's have automated so much of the data scientists workflow that they end up just as you describe, using the tooling and being relegated to becoming analysts.
After 3 years coming back to this field, I see the writing on the wall: In the 90's most models were created by software developers, in the 2030's most models will be created by software developers.
Wouldn’t it have been easier to just use Kafka?
Of course with current cheating the issue is not deep learning based agents but seeing through walls or aim assisting. Those just ruin the enjoyment of the game. Similarly I loathe vs'ing an AI that uses the same methods.
Theoretically at some point it should be possible to make this somewhat permanent with minimal maintenance. Depending on technology it could go from a station that lasts a hundred years to millions of years.