Misplaced Pages

Revision as of 06:05, 20 October 2023

Jimbo welcomes your comments and updates – he has an open door policy. He holds the founder's seat on the Wikimedia Foundation's Board of Trustees. The current trustees occupying "community-selected" seats are Rosiestep, Laurentius, Victoria and Pundit. The Wikimedia Foundation's Lead Manager of Trust and Safety is Jan Eissfeldt.

This page is semi-protected and you will not be able to leave a message here unless you are a registered editor. Instead, you can leave a message here

This is Jimbo Wales's talk page, where you can send them messages and comments.

Put new text under old text. Click here to start a new topic. New to Misplaced Pages? Welcome! Learn to edit; get help. Assume good faith Be polite and avoid personal attacks Be welcoming to newcomers Seek dispute resolution if needed

Archives: Index, Index, A, B, C, D, E, F, G, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252Auto-archiving period: 10 days

This user talk page might be watched by friendly talk page stalkers, which means that someone other than me might reply to your query. Their input is welcome and their help with messages that I cannot reply to quickly is appreciated.

This talkpage has been mentioned by a media organization: Matthew Gault (8 December 2021). "Misplaced Pages Editors Very Mad About Jimmy Wales' NFT of a Misplaced Pages Edit". Vice Media. The trouble began when Wales posted an announcement about the auction on his user talk page—a kind of message board where users communicate directly with each other.

Village pumps

policy

tech

proposals

idea lab

WMF

misc

• Blocks for promotional activity outside of mainspace
• Voluntary RfAs after resignation
• Two-factor authentication for page movers
• Proposed rewrite of WP:BITE
• LLM/chatbot comments in discussions

• view
• edit
• history
• watch
• archive
• talk
• purge

Endowment financial information

Just as an FYI, there's a discussion going on right now to finalize some details and new information will be released very soon (I think before the week is out, barring anything unforeseen). Jimbo Wales (talk) 10:54, 27 September 2023 (UTC)

Financials have been published for the Wikimedia Endowment for the years 2016-2023. Wikimedia Endowment Financial Statements--Jimbo Wales (talk) 17:24, 28 September 2023 (UTC)

Thank you very much for this!

What are our chances of seeing the historical investment holdings? That spreadsheet table is flawed because it assumes half the year's contributions come in at the beginning of the year and half come in at the end, but if we had historical investment transactions it wouldn't be necessary to make such absurd assumptions. Sandizer (talk) 03:55, 3 October 2023 (UTC)

Interesting. Thanks @Sandizer for posting the chart.

I would also be interested in the historical investment holdings. One might guess that Tides/WMF holds, say, treasury bills, treasury bonds, corporate bonds, and so on, generally adopting a more risk-averse investment strategy than 100% investment in the S&P, but in the S&P's worst year (June 2021-22), the WMF's investments did significantly worse than the S&P. I'd like to know what accounts for that. Pecopteris (talk) 05:05, 3 October 2023 (UTC)

I can't answer you, today, as to whether or not we can release that information, because I don't know. I also note that since it's really only at this point a matter of historical interest, I don't think it's worth a huge amount of staff time. I'll check whether it's something easy but I really don't want anyone mucking around with huge spreadsheets to try to answer the question. Leave it with me. It might well be that the balanced answer will be just to get some "color" on the question: "we were invested more heavily in X sector, which performed worse during Y year" or that sort of thing.

However, I can say that going forward, now that we have our own 501(c)(3) I will advocate for regular and clear transparency much greater than we have been able to provide in the past.--Jimbo Wales (talk) 08:37, 3 October 2023 (UTC)

No problem. Here are some typical endowment asset allocations and returns for nonprofit endowments from NACUBO (the National Association of College and University Business Officers) which you might want to join, if they can do such reports at less cost than in-house. Sandizer (talk) 09:16, 3 October 2023 (UTC)

This should also answer the question pretty well: Wikimedia Endowment Investment Guidelines.--Jimbo Wales (talk) 17:19, 3 October 2023 (UTC)

I guess the big difference between the Endowment's asset allocations (the link Jimbo posted) and typical endowment asset allocations (the link Sandizer posted) might account for the big difference between the Endowment's performance (net loss) and the average endowment performance (net gain).

I'm surprised, Jimbo. The Endowment lost millions of dollars -- it failed to keep pace with inflation -- and yet you don't think it's worth investing staff time to find out why, because it happened in the past and is thus of only historical interest? You know what they say about those who fail to learn from history... Levivich (talk) 22:20, 3 October 2023 (UTC)

I should apologise for any ambiguity, as I of course agree with you. The point is, there already is serious analysis internally and I will tell you what I can, but I'm not going to ask staff to spend days of time on specific requests for spreadsheets for people asking questions on my talk page. See below for some information that I think you will find helpful.--Jimbo Wales (talk) 18:55, 5 October 2023 (UTC)

I think not asking staff to spend a lot of time on specific requests from people on your talk page is very reasonable. One staffer recently discussed some of the challenges facing WMF staff when fielding questions from online volunteers, which I'm sympathetic with.

OK so here comes the "but". But, where do we ask? Who do we ask? More to the point, why do we need to ask at all?

"How do the annual returns of the Wikimedia Endowment compare to the annual returns of similar endowments over the same period of time?" is a question so simple and obvious that neither I nor anyone else should have to ask it. It should already be answered here or here or here or somewhere very easily accessible. You are listed on the Endowment's website as the Chair of the Board of the Endowment. I think it's reasonable to ask you, individually, that question -- to ask you to see to it that a public answer to that question is given every year. And if you have to use up staff time to get an answer... well, let's be honest, you've already done that, you already know what the answer is, because the Endowment obviously has already answered this internally, each year.

So that leaves online volunteers like me knowing that while you know the answer to the obvious question, you haven't made it public. ("You" meaning both the Endowment as an organization, and you individually as Chair of its Board). Why aren't you making it public? I know the answer to that, too: because the answer makes the Endowment look bad. It's easy enough to do the math: 12.5% return over six years (2016-2022) = 12.5/6 = 2.08% average annual return, which is less than inflation (3% average over the same period). From this simple number I can figure out that the Endowment performed worse than similar endowments over the same period of time. I might be wrong about that, and I don't know how much worse, but I'd be interested in an answer to that question. An answer that treats me like an adult who knows how to read a P&L. An answer like the kind of answer that a CEO would give to shareholders if the company's investments were returning 2% per year.

Let me bend your ear a bit more with one more specific example of an obvious and unanswered question. The Endowment's Statement of Activities that was just released has this very obvious question that jumps right off the page. That's the $22 million unrealized loss recorded in 2022. Something like a 20% loss in one year.

"What was the $22 million unrealized loss in 2022?" is another super obvious question. Who do I ask? Why do I even need to ask? There are notes on the document itself explaining other figures, but no note explaining this $22 million loss in one year. No mention of it on the Endowment's website or on WMFwiki that I can see. Communication Strategy 101: get ahead of these things. (And please don't answer by saying something like "accounting adjustment"; explain exactly what the heck went wrong that the Endowment needed to record a previously-unrecorded 20% unrealized loss... either the Endowment was way overinvested in some security that tanked, or it way overvalued some securities in the past... either scenario raises further questions.)

So, takeaways: I just learned the Endowment made average net returns of 2% per year while it was with Tides, including a $22 million unrealized loss in 2022. I hope the WMF issues a public statement disclosing how this performance compares with other similar endowments, and if it's worse (as I suspect), explaining why it was worse, what lessons were learned, what steps were taken to make things better. Obviously, moving the Endowment out of Tides' hands is one of those steps already taken. But there's also the question of how we know we'll do a better job managing the endowment in the future than Tides did in the past.

I really feel that the community would not be so harsh with the WMF if the WMF treated the community like adults -- the way CEOs treat shareholders (forthright answers to tough questions) -- rather than the way teachers treat schoolchildren ("trust us, everything is fine, hush now child"). To be clear, I'm not suggesting that you, personally, treat volunteers like children, but I am saying that is how it feels when the WMF/Tides essentially fails to respond to repeated questions about financial performance specifics for years, and then releases a 6-year summary that shows 2% growth and a 20% unrealized loss in a single year, and then tries to spin it on its website as if these were positives. We all know it's bad news; the WMF should be honest that it's bad news, and be transparent about its understanding of why the bad thing happened and its plan for avoiding it in the future. Levivich (talk) 15:54, 6 October 2023 (UTC)

I believe the Endowment's asset allocations have changed over the years. Let me clarify with this summary table.

 *assuming half of all contributions were realized on the first day of the fiscal year, and the other half on the last day.

I'm not qualified to say whether those returns are outside the typical range for such funds, but I don't think they are. Sandizer (talk) 01:22, 4 October 2023 (UTC)

@Sandizer, thanks for this. I spoke with Lisa Gruwell (who acts as President of the endowment) because I want to be accurate, and she provided me with a clear and simple analysis. A few characteristics of the Wikimedia Endowment make it different from the average Endowment in your data. First, the Endowment’s portfolio has been operating under an ESG mandate while it was housed at the Tides Foundation. ESG funds were especially hard hit in 2022 because fossil fuel stocks were one of the “bright spots” that year and we were not invested in that sector due to our ESG mandate. Now that the Endowment is an independent charity, we will be considering our ESG philosophy and I am sure the discussion will generate a lot of debate. (My own belief is that it is possible to invest ethically with little-to-no impact on returns, but this view is not universally held in finance circles.) Second, the Wikimedia Endowment is only 7 years old and some of the big university endowments in the data you shared have been around for centuries. That means the Wikimedia Endowment has not benefited yet from longer-term strategies. The primary difference between our asset allocations and the data you shared is that the Wikimedia Endowment is not yet invested in alternative investments such as venture capital and private equity. We did not feel these were the right fit for us at this initial stage of our development because they tend to be higher risk and are also longer-term strategies. We needed investments during this initial stage that would provide income and liquidity sooner so that the Endowment could cover its own operational costs faster than returns from alternative investments would typically allow. This strategy has worked well for us as the endowment is already covering its expenses and making grants back to the projects according to our spending policy. As the Endowment grows and matures, we may revisit alternative investments, but they are not right for us in this initial stage. Although the S&P 500 is a ubiquitous reference point, the Endowment’s portfolio was established with its own blended benchmark to measure our more diversified strategy. Historically, the portfolio has tracked the benchmark pretty closely. From inception to June 30, 2023, the Wikimedia Endowment’s return has been 5.38 percent, while the benchmark is at 5.51 percent. Slightly lower than the S&P 500, but in the grand scheme of things, not terrible.--Jimbo Wales (talk) 19:00, 5 October 2023 (UTC)

I agree with you that it is quite possible to beat 12.5% over 6 years while still following ESG investing. As a quick comparison, the S&P 500 (.INX) is up 56% over the past five years, while the S&P 500 ESG Index (SPESG) is up 33% over the last five years. Levivich (talk) 18:22, 6 October 2023 (UTC)

12.5% ($15 million) of the Wikimedia Endowment’s total revenue as of June 30, 2023 came from investment results. 87.5% of the Endowment’s revenue came from contributions (donations). I think you are comparing that number to something very different here.--Lgruwell-WMF (talk) 17:10, 7 October 2023 (UTC)

@Levivich I'm replying to your comment above the table on this page here to create a more streamlined reading experience. I am not familiar with the equation you are using to calculate return.  @Sandizer calculated the mean annual rate of return using estimates and arrived at 5.1 percent. He is using a similar method to how we make the calculation.  With actuals, the average annual rate of return for the Wikimedia Endowment from inception to June 30, 2023 is 5.38 percent, as Jimmy shared. CVirtue (WMF) (talk) 20:37, 6 October 2023 (UTC)

@CVirtue (WMF): "During this time, the Wikimedia Endowment had $15 million in investment results (12.5% of total gross revenue that was under management at Tides from inception to June 30, 2023)."

2016-2023 = 7 years, 12.5/7 = 1.8% average per year. Levivich (talk) 21:30, 6 October 2023 (UTC)

You're going to have to expand on what you mean by that, as I have no idea and have tried to figure it out. That's not how the return of a portfolio is calculated over time at all. As far as I can tell, it's a completely meaningless calculation. Jimbo Wales (talk) 09:53, 7 October 2023 (UTC)

It's just taking what the endowment website reported as the "investment results" over 7 years and dividing it by 7. The calculation is meaningless because the 12.5% figure on the website (quoted above) is meaningless. They may have had $15 million in investment results over 7 years, and that may be 12.5% of the total $120M (15/120=12.5%), but that doesn't mean they received a 12.5% return, either net or annual, even though that's what the language on the website I quoted above implies. What's missing from the website is any other % return figure besides "12.5%", like the average annual returns (and by what method it was calculated). I get (from this discussion) that the simple average annual return is 5% not 2%, but it'd be easier for all of us if the WMF would just post the actual annual returns on the Endowment's website, and anyway, whether it's 2% or 5% doesn't make a big difference (though 5% at least beats inflation I guess) to the two questions I'm raising above (which I hope the WMF will publicly answer). Levivich (talk) 14:29, 7 October 2023 (UTC)

It is very important for us to track what portion of the endowment is investment results vs. contributions. Our spending policy prioritizes capital preservation, meaning we do not want to spend the portion of the endowment that comes from contributions. We keep close track this figure because each year the Endowment’s finance committee will set the spending rate at between 2.5 and 4 percent in order to fund the operations of the Endowment and grants to the Wikimedia Projects. This is a meaningful factor for them to consider when making that decision.Lgruwell-WMF (talk) 18:50, 7 October 2023 (UTC)

wikipedia style platform to crowd source feedback for open source LLM

yann lecun suggests, not sure how to set a permanent link to this:

Human feedback for open source LLMs needs to be crowd-sourced, Misplaced Pages style.

It is the only way for LLMs to become the repository of all human knowledge and cultures.

Who wants to build the platform for this?

if it is wikipedia style, should wikipedia build it? host it? ThurnerRupert (talk) 05:16, 16 October 2023 (UTC)

I suspect LeCun might be responding to the surprising results this past week with Zephyr 7B, a free LLM (on Huggingface) which can work quickly on 2018-era mobile phones, and outperforms Meta's Llama-2 70B, an order of magnitude larger model, and does so solely by removing Llama's alignment conditioning (safety censorship) which has come as quite a shock to many, and is not great news for those who consider alignment-oriented fine tuning important. (I personally am skeptical because while heavily aligned models will refuse to discuss some sensitive topics, they are usually very willing to discuss subsurface details like how to use and acquire the specific equipment for weaponizing anthrax spores, for one particularly egregious and confusing example.)

As to your question, all of the commercial LLM chatbots crowdsource human feedback (the thumbs down/up buttons) but there are no open source LLMs which have integrated this technique yet as far as I know. There are some commercial companies offering it as a paid service and having open sourced the code behind it, but the administration overhead for reintegrating such feedback into the models is far more costly than the code. LeCun on the surface is suggesting the open source LLM community find a way to crowdsouce the reintegration overhead too, which I'm sure is doable, but might not attract volunteers. Sandizer (talk) 10:08, 16 October 2023 (UTC)

P.S., some interesting discussion in the replies to LeCun's tweet on the topic. Sandizer (talk) 00:50, 17 October 2023 (UTC)

Happy October, Jimbo! :)

Babysharkboss2 has given you a Hershey Bar! Hershey bars promote WikiLove through chocolately goodness and hopefully this one has made your day better. Hershey bars are wonderfully delicious! Spread the WikiLove by giving someone else a Hershey bar, whether it be someone you have had disagreements with in the past or a good friend.

Spread the goodness of Hershey bars by adding {{subst:Hershey Bar}} to someone's talk page with a friendly message!

Babysharkboss2 was here!! 15:51, 17 October 2023 (UTC)

Hell broke loose today

Elon musk is calling Misplaced Pages woke u replied Then they start talking about your connection with newsguard is this true I'm still kinda confused about the woke claim when most of us work to be neutral If you would like to clarify your position in newguard I would like to know Elon musk also is a asshat •Cyberwolf• 01:40, 20 October 2023 (UTC)

Useful reading: Punctuation. AndyTheGrump (talk) 02:06, 20 October 2023 (UTC)

huh. hell broke loose today? i thought it did that every day. ltbdl (talk) 02:24, 20 October 2023 (UTC)

NewsGuard? Not exactly hell. Johnuniq (talk) 02:56, 20 October 2023 (UTC)

Yeah, Elon going on a rant isn't really my definition of "hell". It's more my definition of any random day. I am an advisor to Newsguard, yes. I don't think that's in any way problematic. People who may think that I'm somehow in favor of censorship are... just making things up I'm afraid. What I am in favor of is products and services that are well designed to help people evaluate and access high quality information. At least some of the trolls who hate such things hate them because of reasons that should be obvious: Newsguard would tend to rate various crackpot propaganda sites rather less highly than trusted news organizations with standards of fact checking, running corrections, etc. Their nine critera strike me sa being fairly straightforward.--Jimbo Wales (talk) 06:05, 20 October 2023 (UTC)

• Misplaced Pages pages referenced by the press