- Homescreen
- Posts
- š "We have no moat" - Google
š "We have no moat" - Google
Oh yeah neither does OpenAI
Gm. Another weekend, another party full of conversations about AI. š
POLLS
41+ āEach episode of "Sex and the City" inspired me to a new pair of shoes. How many is many of Manolo Blahnik shoes? Kidding me. Never enough. It's quality what counts. And don't get me started on the Dolce & Gabbana shoes...ā You and my fiancĆ©e would be best friends.
6-10 āBesides sneakers, formal shoes, and Crocs why would I need more shoes? It's not that I have more than 2 feet.ā I need to buy some Crocs!
11-20 āThis just made me realize I need to get rid of some shoes šā You are very welcome. Spring cleaning is here.
41+ āroughly 300ā. Thatās almost 1 per day, which is honestly goals. I heard this story that Jay-Z wears a pair of brand new white Air Force Ones every single day and I think itās a myth but tbh sounds like the dream.
11-20 āNeed more dunks and AF1sā Facts, you can never have enough AF1s.
41+ āDear Nike, take my money. That is all!ā Siri, play Nikes on My Feet.
š AI
OpenAIās Regulatory Troubles
If youāre reading this newsletter, chances are that chatGPT has become the robot assistant youāve always wanted.
Unfortunately, not everybody loves our favorite AI chatterbox. ChatGPT has persistently been in hot water with regulators, and that is true in no place more than Europe.
The GDPR
The issues for ChatGPT stem from Europeās adherence to the General Data Protection Regulation (GDPR), one of the worldās strongest data privacy laws.
It was ChatGPTās alleged breaking of the GDPR that led to Italy giving it the boot, and although the app has since returned to the land of pizza and pasta, Germany, France, and Spain have since launched their own probes into OpenAI and ChatGPT.
A Variety Of Concerns
The regulatorsā concerns really boil down to how ChatGPT uses peopleās data. And, because ChatGPT is trained on an immense data set that most people have no idea they are even a part of, the app runs smack into the GDPR, as:
It requires companies to have explicit consent before collecting and using personal data.
It requires transparency over how data is used and stored, something OpenAI isnāt really doing.
It requires users to be able to demand that companies remove their data, something that might not even be possible in large language models.
It requires that all personal data must be accurate, a problem for the current hallucinating-prone chatbots.
It bans collecting data from those under 13, something that OpenAI does not currently monitor for.
Thankfully, it appears that OpenAI is aware of Europeās data concerns, as they recently rolled out the ability to turn off your ChatGPT chat history (providing an opt-out of providing training data) and announced that it would stop training on its customersā API data.
Theyāve got a long way to go to get the regulators off their backs for good, but itās a great start.
Why should I care? Data privacy laws may be annoying to the AI companies, but they are important to us consumers. If OpenAI and its peers are able to follow the regulations, then we will have AI tools that donāt infringe on our privacy. What a world thatād be.
š AI
āWe have no moatā - Google
In a leaked document from inside Google, a researcher was quoted saying āWe have no moat, and neither does OpenAI.ā
A moat is a competitive advantage ā like a moat around a castle. And if an employees says his or her company has no moat, well thatās pretty bad. And if an employee says even their competitor doesnāt have a moat, thatās even worse. So who has the moat?
āThe uncomfortable truth is, we arenāt positioned to win this arms race and neither is OpenAI. While weāve been squabbling, a third faction has been quietly eating our lunch. Iām talking, of course, about open source. Plainly put, they are lapping us.ā
The researcher then gave a list of āmajor open problemsā that the open-source AI community is solving:
LLMs on a Phone: People are running foundation models on a Pixel 6 at 5 tokens / sec.
Scalable Personal AI: You can fine-tune a personalized AI on your laptop in an evening.
Multimodality: The current multimodal ScienceQA SOTA was trained in an hour.
āOpen-source models are faster, more customizable, more private, and pound-for-pound more capable. They are doing things with $100 and 13B params that we struggle with at $10M and 540B. And they are doing so in weeks, not months.ā
Implications according to the researcher include:
āWe have no secret sauce.ā The researcher believes Googleās best bet is to learn from and collaborate with people outside of Google.
āPeople will not pay for a restricted model when free, unrestricted alternatives are comparable in quality.ā The researcher believes they need to consider where their value really is.
āGiant models are slowing us down.ā The researcher believes the best models are the ones that can be iterated upon quickly.
Well, if youāre at Google or OpenAI, this may be slightly terrifying.
Why should I care? Itās a good time to be an open-source dev. The researcher says āwe need them more than they need us.ā The alpha may no longer be coming from Google and OpenAI, but from the builders on the outside.
šāāļø QUICKIES
What else is going on?
TikTok kept track of people who watched gay content.
Joe Biden isnāt allowed to join the cool kids on Bluesky.
We now know whatās inside the moon. And itās not cheese!
AI is really freaking good at designing mRNA vaccines.
Stat: 18%: How much Coinbase beat revenue expectations by ($773M vs. $655M expected). Inject a crypto bull run into my veins.
Rabbit hole: Seismic Waves Of Gen Z Behavior (Digital Native)
š¤©Ā MONDAY MOTIVATION
How to Stand Out via Garry Tan:
š ļø FOUNDERS CORNER
The best resources we came across this weekend that will help you become a better founder, builder, or investor.
šŗ Cartographer is a tool for making market maps. VCs can now save approximately 4 million hours per year.
šø Venturefy AI is a public wiki of verified B2B relationships.
š» DevDynamics is a new tool that gives you analytics of your code velocity. How fast are you shipping and deploying code?
š DOPAMINE HIT
When You Watch Too Much Succession
HOW WAS TODAY'S NEWSLETTER? |
REACH 30K+ FOUNDERS, INVESTORS & OPERATORS
If youāre interested in advertising with us, send an email over to [email protected] with the subject āHomescreen Adsā.