kozikow

> For start, day trading is effectively gambling.

Not unless you managed to get hired as a (quant) trader with a performance fee. Privatized gains, shared losses. Of course, it's very hard to get hired for such a role.

I'm currently doing BI evaluation for improvement of the current toolkit, so I might use this comment section for some BI experts' advice.

We mostly use Looker Studio + Metabase currently.

Investigating Tableau, Looker, PowerBI, Sisense, Omni currently. PowerBI is great, but BigQuery integration seems to be lagging. Is there a way to get such a tool without paying $50K a year? Everything so far is only a "nice to have" upgrade over our mostly free toolkit. We also care about good geospatial visualizations and embedding reports for users without requiring a login.

Where GPT really improved for me subjectively is long context - at least since launch. Such ranking can't compare it.

Also I felt I keep getting more personalized results, i.e. models are somehow biased towards user. I heard they plan it, I don't know it's launched, but I feel it.

And there's also fine-tuning in the other direction - my brain got used to ways of interacting with GPT. Same as with Google, I just somehow subconsciously know how to write prompts that get me what I want.

All of those variables they described are a contention point in other coffee-making methods like espresso as well.

And many of those options are "correct", but depend on your preference - do you like more concentrated coffee? Do you like it less bitter? Do you want more caffeine? Etc.

I think many founders have some buried deep worry like this.

You end up fundraising, preparing forecasts and budgets for investors for the next X years, attending conferences, working with marketing on PR, dealing with unglamorous problems that are not part of anyone's job but need to get done and someone else doing it would impact morale too much. It's lucky if you find 1-2 days a week for "deep work" given constant context switching.

Even if you work 2 shifts you can't sometimes match the level of being necessary for day-to-day operations as "regular" employees. But you own orders of magnitude more equity.

I'm more interested how those geospatial visualization were done. Do you know a probable library used for it?

I think what should be mentioned in the article is bridge rounds. I know various startups that have done it. In my feeling, at least half of the 38% would be explained by VCs raising bridge rounds for their existing portfolio companies.

Rather than come clean that all your valuations came down by 2-3x (the reality of what valuations are circulating on the market nowadays) just do things like raise convertible note with a cap of last round valuation and don't announce it on pitchbook/crunchbase.

It's not easy for VCs to just pause investing or invest in some other stuff than their "thesis" - some have agreements legally binding on what they would invest in and timeframes for spending the capital.

Just because the acquired company falls into the "can't justify higher rev multiple thanks to synergies with the acquiring company" and "not successful enough for the IPO".

If there was no options in that niche, the market objectively would be a lot tougher.

There's also always a private equity market that is often a middle ground between IPO and acquisition.

And in my feeling, I've seen PE getting more active in tech last year (maybe just because valuations went down).

My view is to focus on doing stuff. That's what I did. Pick up some task you want the model to do, try finetuning llama, playing with APIs from OpenAI, etc. Googling and asking GPT along the way.

Foundational model training got so expensive that unless you can get hired by "owns nuclear power plant of GPUs" you are not going to get any "research" done. And as the area got white-hot those companies have more available talent than hardware nowadays. So just getting into the practitioner area is the best way to get productive with those models. And you improve as a practitioner by practicing, not by reading papers.

If you're at the computer, your time is best spent writing code and interacting with those models in my opinion. If you cannot (e.g. commute) I listen to some stuff (e.g. https://www.youtube.com/watch?v=zjkBMFhNj_g - Anything from Karpathy on youtube, or https://www.youtube.com/@YannicKilcher channel).

Those systems when they work are OK.

But connectivity is so annoying. 1/4 times Android Auto fails to connect. I have to restart the phone, restart the car and it connects. And I'm not surprised it's all ducktape security.

Or another way to put it - if you are an enterprise based in Europe that needs to stay compliant, future regulation will make it very hard to not use Mistral :P.

Not OP, but For me:

- Writing: emails, documentation, marketing - Write a bunch of unstructured skeleton of information. Add a prompt about the intended audience and a purpose. Possibly ask it to add some detail.

- Coding: Especially things like "Is there a method for this in this library" - a lot quicker than browsing through documentation. Some errors - copy-paste the error from the console, maybe a little bit for context, and quite often I get the solution.

And API based:

- Support bot

- Prompt engineering of some text models that normally would require labeling, training, and evaluation for weeks or months. A couple of use cases - unstructured text as an input + prompt, JSON as an output.

It's funny because now the OpenAI keynote feels like it's emulating the Google keynotes from 5 years ago.

Google Keynote feels like it's emulating the Apple keynote from 5 years ago.

And the Apple keynote looks like robots just out of an uncanny valley pretending to be humans - just like keynotes might look in 5 years, but actually made by AI. Apple is always ahead of the curve in keynote trends.

It's the least defined role.

Currently, I am in a funny situation when all teams agree we need an additional data engineer. But basically:

- Sales and finance want more of business intelligence analyst

- Devs want more of a backend engineer

- ML researchers want a data analyst proficient in ETL to do pipelines on the training dataset

All of those 3 have only one thing in common - they need to know SQL very well. I've worked extensively with various technologies to analyze data - pandas, sql, spark. And still, I find SQL (especially recently BigQuery) getting me what I want the quickest.

> how they only want the best candidates and how hard their interviewing process is

I sometimes feel like those algorithmic challenges at interviews are more for making developers feel they work at the "top team" than filtering the candidates. I am not salty about it - I am very good at those, did lots of Topcoder back in the day.

Most of the people in the industry agree that they have a terrible signal-to-noise ratio (or any other standardized interview method for developers). Any method that works doesn't scale, but large IT corps don't want to hear that.

I don't think it's too far-fetched. To see where it's going look at social score in China.

You say something wrong about a party, suddenly you can't board a plane, take a mortgage, enter some buildings, ...

Your credit score would look at how compliant you are with policies that can get increasingly nonsensical.

Surprised it's just 18%.

2021 was a height of "more money on the market than productive things to do with it". Startups are expected to operate in 2-year fundraising cycles - VCs used to push you to increase burn rate if you had a longer runway.... So right now startups that raised at the top of 2021 are facing potential fundraising now. The equivalent company would have 2-3x lower valuation in 2023 compared to 2021 - based on my discussions with other founders, what board members say, etc.

That "unfortunate" probably has been deliberately put there by Sam, possibly for the reason you said.

To be honest I extracted only one useful piece of information here.

"No particular comment on that unfortunate leak"

The word unfortunate for me sounds significant. If it was random noise as so much we heard recently, such a word wouldn't be used and so much emphasis wouldn't be given to this paragraph, right?

The progress from GPT3 to GPT4 has been so substantial that many might argue it signifies the advent of Artificial General Intelligence (AGI). The capabilities of GPT4 often elicit a sense of disbelief, making it hard to accept that it's merely generating the most likely text based on the preceding content.

Looking ahead, I anticipate that in the next 2-3 years, we won't witness a sudden, magical emergence of "AGI" or a singularity event. Instead, there will likely be ongoing debates surrounding the successive versions like GPT5, GPT6, and so on, about whether they truly represent AGI or not.

Everyone is still left guessing what have happened.

At least my speculative view (rewritten by chatgpt ofc):

There's speculation that OpenAI's newly developed Q* algorithm marks a substantial advancement over earlier models, demonstrating capabilities like solving graduate-level mathematics. This led to discussions about whether it should be classified as Artificial General Intelligence (AGI), a designation that could have significant repercussions, such as potentially limiting Microsoft's access to the model. The board's accusation of "not consistently candid" against Sam Altman is thought to be connected to efforts to avoid this AGI classification for the model. While the Q* algorithm is acknowledged as highly advanced and impressive, it's believed that there's still a considerable journey ahead towards reaching a true singularity event. The complexity of defining AGI status in AI should be noted. At what point of advancement can a model be labeled as an AGI? We are way past simple measures like the Turing Test. Arguments from both perspectives have their merits.

I read the news, make a picture of what is likely happening in my head, and every few hours new news comes up that makes me go: "Wait, WTF?".

We don't know the structure of their new unit, do we? Sometimes "startup in a big corp" may really bring the best of both worlds (although in reality, 90% of such initiatives bring the worst of the two worlds).

For many years, Microsoft Research had a reputation for giving researchers the most freedom. Probably even that's the reason why it hasn't been as successful as other bigcorp research labs.

I know Jakub Pachocki and in my feeling, Jakub leaving is worse for gpt than Greg or Sam.

Im not claiming to know more than everyone else, but Sam was IMO just a face. Greg is backend engineer - that's less important than actual research.

My prediction : Sam agrees to some business deal the rest of the board found unethical. Something of magnitude "chatgpt used by the NSA".

Other predictions I've read here sound unconvincing to me. Already developed AGI, spending too much on compute - this is not something CEO alone could hide and I would trust other board members to understand those matter better.

That's not just external facing problem. Even when internally you want to figure out something - e.g. get a quota for something it's very hard to locate who could get things done for you.

There is this attitude that humans shouldn't ever talk to other humans, but to systems. So if your use case is not handled by how the system was designed, talking to humans would take many months, so better just give up.

The problem may be even worse internally tbh. You want to use some project, there's old teams page, one email, you email person there and after 3 pings weeks later they reply they work now on a different team.

That would be very easy to circumvent.

You can describe effectively the same software engineering job in a way that sounds completely different. "close a project" of the old team, etc.

If you go to any startup founders gathering, I typically wonder "I am usually the biggest weirdo in the room. But in this crowd I feel among the most normal people ".

And it's still true for the later stage startups as well, but indeed it's the most prominent at the seed stage. There's really just a lot of weirdos among startup founders.

Tech saying "I'll become farmer" to escape tech is very amusing, especially how data-driven the large scale farming is, and how small scale farming is getting priced out by large scale farming.

I'm not a farmer, but I did some tech for farming and you would be surprised how tech driven it is. Agriculture was probably the first industry that used satellite imagery outside of military on the large scale.

If someone is Silicon Valley web app developer and went farming, they actually could be going deeper into tech than escape it.