Vespa.ai is spinning out of Yahoo as a separate company
blog.vespa.ai
blog.vespa.ai
Even though I'm with Zilliz/Milvus now, I wholeheartedly support and recommend folks check out and try Vespa. Congrats to the Vespa team!
EDIT: For folks on Twitter, you should follow Jo (https://twitter.com/jobergum) from Vespa if you aren't already. Great combo of technical content, hot takes, and vector database memes!
(Fwiw I have Twitter blocked too, though not for moral reasons)
The subtext is clearly “jkb, please move to Mastodon cause I blocked Twitter”. Not spelling that out doesn’t make it a lot less weird IMO.
I am learning lsh right now and find it fascinating
I've been meaning to dive into SCaNN and DiskANN as well, but haven't gotten around to it yet.
quick TLDR on both vs HNSW?
Having worked with both Solr and Elastic in past search companies, it’s incredible to be able to deploy into Fortune 50 enterprises without any doubts about stability and with all the benefits of a cutting-edge hybrid engine.
Can’t wait to watch you on the next leg of your journey!
The Atolio team
what's usually going on in scenarios like this is either: the subsidiary is in growth mode and needs access to public capital markets to raise large amounts for building infrastructure (and sharing the expenses with others) to compete, or what has been built is going into profitability mode and the profits are better protected/hoarded in a private company where disclosure requirements are not onerous. If this move represents the former, expect the latter in a few years, is why yahoo would maintain a large stake.
(I'm just pointing this out because there are frequently underlying reasons that have nothing to do with the type of stuff that's in the press release, "successful venture, income/exit event" etc.)
So yahoo spins them out and has their equity grow faster than yahoo stock, or they lose the product to a new competitor.
A "spin-in" is a form of R&D in which a company is the sole investor in a startup. It sends a team of employees off to build an experimental product and then buys that startup for a predetermined and very healthy price.
Cisco has spent on average $763 million to acquire each spin-in that these three men have founded, even though Cisco was also the sole investor in each.
https://www.businessinsider.in/why-cisco-has-showered-these-...
I don't know about you, but that seems likely to be an excessively inflated figure to me (a la the dot-com bubble.)
Comparing any company's value to its valuation during the dot-com bubble seems silly to me.
Functionally speaking when a company issues a dividend of $x the stock price S will drop to S-x on the ex dividend date. This makes sense as the company now has $x less per share. The shareholder is at a wash as well. They had S before and now they have S-x+x. No value was created, money was simply moved.
Given the above you can more or less value dividends using discounted cash flow. A ~3% dividend is CSCO’s case isn’t worth paying more (increasing the share price over) on its own because the cash used to buy the shares is worth ~5.5% today according to the risk free rate.
In fact if one were considering the dividend alone then the company should be worth less. Obviously factoring in future growth, etc… the market has landed on the current price.
This will be wrong but true enough to make my point; With a market cap today of ~$215 billion, had CSCO not paid out dividends, their market cap would be $215 billion that represents current cash and the discounted value of future cash + the $54 billion that would also be in cash -- for a total of $269 billion. That's enough to add ~$15 to their share price.
It seems like the smart move to spin it out to access this money.
Vespa's been a technology built with search & ad-serving in mind, while Yahoo's other recent moves have been ad-tech layoffs alongside acquisitions in sports, finance & tech journalism, so the focus shift makes some sense here.
I remember understanding it as the internal equivalent to something like Solr, although actually it predated it - looks like the first public Solr open source release (coming out of CNET) was January 2006.
Back then, the properties had to stand up their own Vespa cluster(s), later on we created a managed service out of it. And, yes, the original plan for Vespa was to be a Vertical Search Platform, that is where the name Vespa comes from. More on the history in this blog post https://blog.vespa.ai/vespa-is-becoming-its-own-company/
True. I bought my first Vespa in 1984
Vespa really is an impressive platform. I've been working with search platforms for a while, starting with FAST Instream, MarkLogic, Elastic, Weaviate, and now Vespa. As others have noted, Vespa has a deep technical lineage. You can see a lot of the hard earned lessons in the design, as well as improvements in the flaws of the spiritual predecessors (FAST et. al).
The history alone makes the project interesting. But then, in the last couple of years (or maybe earlier), they started introducing all the fundamentals for supporting deep integration of tensors and a custom HNSW implementation. Whether accident or planning, this was a nice strategic move. They were leagues ahead, just in time for the popularization of learned embeddings.
Now they have the best combination of traditional lexical search, semantic search with embeddings, and the combination of the two in hybrid search. That's without even getting to all the functionality in the multi-phase ranking, hosted ML models, and document processing engine.
Obviously I'm biased (we use Vespa at work for enterprise search products), but I can't recommend this engine/platform enough.
Congrats again to the team!
https://blog.vespa.ai/vespa-hybrid-billion-scale-vector-sear...
This blog post says it's "developed by Yahoo" which is I guess true. But it was originally an acquisition, largely developed by a team in Norway, and apparently most Vespa development still happens there.
very, it's outcompeting most vector databases on features and maturity when it comes to vector search while having very powerful and flexible and proven text search too
The tricky part is, it's more a platform to build complex search systems with then "just" a vector database. So if a found a company today which focus is to create clever multi phase search pipelines and train (e.g. domain adopt) LLMs for calculating embeddings etc. then it's probably _the_ best solution by far, you probably can get away with having only AI engines devops and a single programmer (who might also most times just do devops). But if you need to deeply integrate it into a different existing search system things are less grate.
And I would love to see some modernization, like having a format for structured queries which is more widely supported then YQL... (eying graphql here)
Feel free to explore our open-source benchmarking tool, which allows you to examine our methodology and even compare it with your vector database. https://github.com/zilliztech/VectorDBBench
https://upload.wikimedia.org/wikipedia/commons/e/e6/Polar_Be...
So a few months later I woke up early one morning and saw it had been snowing all night. I made sure to get over to the office before the traffic started ruining the image of barren, desolate place where you could entertain the idea of polar bears roaming the streets, looking for young, fresh engineering meat.
I took a few photos, walked inside, and got busy looking for photos of polar bears that could fit into this idea of upping the Dangerous Wildlife War between India and the home of the proper vikings.
So I did a little bit of photoshopping, uploaded it to my Flickr account, and sent the link to a few other yahoos around the world. We had a bit of fun with it, but I don't think anyone with access to this worldwide web of information will be too fooled. (It's quite easy to figure out that there are no polar bears on the Norwegian mainland, so I've never been too bothered about leaving it up there.)
You can see the ehm... original at https://www.flickr.com/photos/eirikref/3262204184 for a few of the comments and fun from back then.
Edit: So that's about as definite as you'll ever get it. And I'll never again admit to it being photoshopped in any shape or form ;)
Was there an other side to this war? The most dangerous animal you're likely to come across (realistically, without getting photoshop involved) in an average Indian street would be a monkey. I've also come across many varieties of snakes, from time to time, but they're generally chill and just want to be left alone. The monkeys are ruthless bastards though, not to be trifled with.
The only ones to ever get in touch about this whole thing was, 1) The Y! corporate blog who wanted us to make a lighthearted a fun write-up about the polar bears, which I cannot seem to find online anymore, and 2) a random local from Trondheim, Norway who sent me quite the angry e-mail telling me that I was single-handedly ruining the tourism business in town with that silly image.
Don't sleep on some of its newer features like multi-vector document fields, either…
It is believed that Apple paid the Beatles a little over $500 million for Apple Corp trademark rights so they could market “Apple Music”
https://en.m.wikipedia.org/wiki/Apple_Corps_v_Apple_Computer
The Vespa is an iconic and well known brand in Italy and worldwide. If Piaggio isn't litigating it, I wonder what their lawyers are even doing.
As to the former, trademark classes are one of the ways in which confusion is determined, but it’s not the only way for sure.
That said, if it’s a word borrowed from another language and there’s no intent to deceive, I imagine courts would allow it - see also Apple/Apple Corps, and the many companies called “Delta” not in the airline business.
If you look for just pure vector similarity search, there are many alternatives. But Vespa's tensor support, multi-vector indexing and the ability to express models like colBERT (1) or cross-encoders makes it stand out if you need to move beyond pure vector search support.
Plus, for RAG use cases, it's a full blown text search engine as well, allowing hybrid ranking combinations. Also with many pure vector databases like Pinecone, you cannot describe an object with more than one vector, if you have different vector models for the object, you need different indexes, and then duplicate metadata across those indexes (if you need filtering + vector search).
1 https://blog.vespa.ai/pretrained-transformer-language-models...
Not to mention the naming trend companies took on for a decade afterwards.
What is perhaps a bit unusual is that the Vespa team has existed for a very long time. Even before the project got its name (2004/2005?). If memory serves it will be 20 years this year or next year since the first bits of what became the Vespa platform were developed. Some of the people working there have worked together since 1998-1999.
I think much of the reason for the longevity of the team is that it was kept somewhat separate from the rest of Yahoo! so they could maintain focus.
It definitely has legs and if the team understands just how enormous the potential could be, annd especially why they could disrupt, I’m even more eager to see what they do.
ZHoldings is now Line Yahoo.
The footer of yahoo.co.jp has changed from Yahoo! Japan / ZHoldings to Line Yahoo.
My point is just, from what I've heard the switch to Line Yahoo reflects a real integration of the businesses, not just a change of brand.
Yahoo is a profitable company with atleast 5B $ in revenue. Search and Mail are extremely profitable businesses. When was the last time you changed your mail id? Search is atleast 2B $ plus in revenue. Mail is 1B $ plus in revenue. Yahoo Finance, Yahoo sports, yahoo news are all highly trafficked websites although I dont know the revenue numbers. I do wager that if yahoo is put on stock market, its valuation will be around $10B although private equity has spent only 4B$ on acquiring them. Who knows what went into the discussions.
Why did I leave? Because I figured my career growth is elsewhere.
One thing I learnt at yahoo is that large companies take a very long time to die . You can have a profitable career in them, even if YoY growth is slowing.
(I worked on Yahoo Search — and search.yahoo.com — from 2007 to 2010)
I get this sentiment, but I don't know anyone in my age group (late 20s/early 30s) that use yahoo mail either. My dad did use yahoo mail, but i'm guessing it has about 20 years left.
In 2006, when I moved from Yahoo Mail to Gmail.
19 years ago, the minute gmail offered a better alternative to Yahoo!
https://yourstory.com/2023/05/yahoos-billion-dollar-blunders....
In the same year that Google had Borg, and AWS released EC2, at Yahoo, David Filo was still personally reviewing hardware requests.
People tend to forget or are to young to know just how big and innovative Yahoo was at one time. I know this may sound strange if you're younger but Yahoo was literally the Google of their era. Yahoo was absolutely *massive* at their height.
The technology of yesteryear, I miss it so.
The biggest thing I recall about Yahoo is Jerry Yang vetoing sales to Microsoft for $35B on principle because he didn't like Microsoft.
Is that what actually happened or is that a false memory?
I really don't know how Yahoo managed to fumble things so badly but they were always one of the better companies as far as technology goes - this announcement today is a testament to that. Just that their business and product side was lacking
"The deal for Alibaba was expensive and risky at the time, but ended up as the most lucrative bet in Silicon Valley: at today's prices, that stake would be worth more than $80 billion. Normally, such a success would preserve a company for years, but on Monday morning Yahoo announced that it would sell its operating business—the web properties without Alibaba and other investments—to Verizon for less than $5 billion, pennies on the dollar it would have commanded at its peak 15 years ago."
Like, when naming things...maybe not use a name that people use to define a category. (AKA: A Vespa is to Scooters as a PostIt is to sticky notes)