DEV & SEO Agency Owner // way harder than Neo in the first matrix ☠#ranksanta👻SEO, Full Stack Dev, AI Quant Fintech SASS block ⛓️⏳#aspie / #ADHD
Gold Coast, Queensland, AU
Joined June 2014
- Tweets14.5K
- Following10.4K
- Followers17.8K
- Likes17.2K
Aaron Whyte ♠️💾 retweeted
"Professor, don't you find it curious that a new US-Iran peace deal leaks almost every time the 10y UST yield breaks 4.4% on the upside?"
"Actually, if I think about it, I don't find it curious at all."
Aaron Whyte ♠️💾 retweeted
Hormuz is being modeled by consensus as an oil shock. It is not. It is an input layer reset of the global manufacturing economy. The crude price move is the loudest signal but not the most consequential.
The most consequential transmission is the slow repricing of the chemical, metal, and specialty input layer that sits underneath every physical product made on Earth.
Each cascade has its own time signature. Oil moves in days. Fertilizer in weeks. Specialty chemicals in months. Capital goods and consumer durables in quarters. Sovereign wealth flows and reinsurance capital in years.
So the impact rolls through markets in waves rather than a single shock. This is what makes it harder to model than a typical commodity event.
Aaron Whyte ♠️💾 retweeted
The terrible takes on oil are killing me slowly.
Now may be a good time to familiarize yourself with this. Not all oil is equal, and different grades produce different products.
This is super cool! >>Platts periodic table of oil...it is interactive! Click on the cells in the link to find out more about each one
spglobal.com/platts/plattsco…
Fun with #telstra yet again.. landing in a different country to find they've updated their roaming and app mid flight and now Ur locked out with no access to calls, txt or data
Aaron Whyte ♠️💾 retweeted
Brew some coffee and join us for @TheMarketHuddle with @PatrickCeresna and @capitalistexp
This week, I stand in for @kevinmuir and pretend I understand Crayon Voodoo with Patrick…👇👇
.@hkuppy steps into @kevinmuir's shoes as @patrickceresna's co-host this week. Pat and Kuppy welcome back to the show, Capitalist Exploits founder, Chris MacIntosh @capitalistexp. The guys discuss bonds, rates, economy, shippers, offshore oil and of course, Uranium.
Aaron Whyte ♠️💾 retweeted
Only reason it rallied is because RBA might hike, otherwise would’ve fallen below 62, it did reach 62.7 at one point. Do not underestimate the probability of a currency crisis if RBA doesn’t hike, UK had done last year (our real interest rates are deeply negative) #ausbiz #forex
Do u really think u hav that much control over your adult children lol 😆 – at Gold Coast, Queensland
Aaron Whyte ♠️💾 retweeted
"You can measure everything about a bubble except the most important part: When investors will stop believing in it. The end of the bubble is just the end of enthusiasm. And enthusiasm isn’t a tamable statistic. It’s a hormone that owes nothing to the logic of your data." @morganhousel
Here we go, again – at Gold Coast, Queensland
Three British news organisations have reported that comedian Russell Brand has been accused of rape, sexual assault and abuse based on allegations from four women at the height of his fame.
#9News nine.social/lBn
Aaron Whyte ♠️💾 retweeted
A tick-up in loan arrears at non-bank lenders has caught the eye of the prudential watchdog #ausbiz afr.com/companies/financial-…
Aaron Whyte ♠️💾 retweeted
Loans to various businesses, resulting in investments in properties across Box Hill, Toongabbie North Kellyville, Austral, Oren park & Wentworthville were targeted for development. @_mumbling_me @Ryan79728737 @sachinsyd
We can tots trust their forecasts – at Gold Coast, Queensland
The RBA's updated and extended forecasts show underlying inflation falling from 5.9% now to 2.8% by Q4 2025, just a little inside the target band but above the 2.5% recommended by the RBA review, assuming the cash rate peaks at about 4.25% and falls to about 3.25% by end-2025. Previously the RBA forecast inflation would fall to 2.9% by Q2 2025 assuming cash peaked at around 3.75% and fell to about 3%.
Aaron Whyte ♠️💾 retweeted
INFLATION: #accelerates +10% since Old Wall declared victory on "fighting inflation"
Aaron Whyte ♠️💾 retweeted
GPT-4 is getting worse over time, not better.
Many people have reported noticing a significant degradation in the quality of the model responses, but so far, it was all anecdotal.
But now we know.
At least one study shows how the June version of GPT-4 is objectively worse than the version released in March on a few tasks.
The team evaluated the models using a dataset of 500 problems where the models had to figure out whether a given integer was prime. In March, GPT-4 answered correctly 488 of these questions. In June, it only got 12 correct answers.
From 97.6% success rate down to 2.4%!
But it gets worse!
The team used Chain-of-Thought to help the model reason:
"Is 17077 a prime number? Think step by step."
Chain-of-Thought is a popular technique that significantly improves answers. Unfortunately, the latest version of GPT-4 did not generate intermediate steps and instead answered incorrectly with a simple "No."
Code generation has also gotten worse.
The team built a dataset with 50 easy problems from LeetCode and measured how many GPT-4 answers ran without any changes.
The March version succeeded in 52% of the problems, but this dropped to a pale 10% using the model from June.
Why is this happening?
We assume that OpenAI pushes changes continuously, but we don't know how the process works and how they evaluate whether the models are improving or regressing.
Rumors suggest they are using several smaller and specialized GPT-4 models that act similarly to a large model but are less expensive to run. When a user asks a question, the system decides which model to send the query to.
Cheaper and faster, but could this new approach be the problem behind the degradation in quality?
In my opinion, this is a red flag for anyone building applications that rely on GPT-4. Having the behavior of an LLM change over time is not acceptable.
Have you noticed any issues when using GPT-4 and ChatGPT lately? Do you think these problems are overblown?
ALT Accuracy comparison between the March version of GPT-4 with the June version on the problem of determining whether a number is prime. In March, GPT-4 solved 97.6% of problems accurately, while in June, it solved only 2.4% of the problems.
