Filip Fors Connolly

@klarnaseb @ken_wilson__ Innovating much slower? That was the feeling before, but not anymore. Google's trend this year is actually more impressive than OpenAI's

@ken_wilson__ I am taking about where the trend is going. Not what is true today. They are innovating much slower. I try their stuff now and then but it is just not that much better it changes my habit.

Stock markets always baffle me. Tomorrow Google stock will rebound due to tariffs. Meanwhile OpenAI just announced 1 bn users? How is this not the beginning of the end for Google? Personally I barely use it. Why is that not a shock event to the market?

@johnrushx The jump from GPT-3 to GPT-4 was likely a much larger increase in compute compared to GPT-4 to GPT-4.5. Improvements from compute are following a logarithmic scale.

We hit the wall with LLMs > Two years ago, GPT4 made AI wrappers & Agents possible. > Sonnet 3.5 solved the AI code generation. > I just tried Sonnet3.7, GPT-4.5, O-3, Grok, and Gemini... > Not impressive. Not even close to gpt3->gpt4 jump What's next:

@UniverseIce Do you think the front screen on Mix Fold 4 will be wider than the previous models? Prefer Pixel Fold over Mix Fold 3 due to the superior aspect ratio on the front screen.

Honor Magic V3 will be released on July 12th. Then, Xiaomi MIX Fold4 will also be released, which is also the narrowest foldable phone in the world, only 9.x mm. It is worth looking forward to that MIX Fold4 will be released in the global market! You have a chance to buy it!

@rmwillen Gratulerar!

@frech_johannes Milen under 40 och att kunna dunka en basketboll saknas.

När eleven samråder med handledare kan hon eller han med säkerhet bedöma den egna förmågan och situationens krav. Dessutom kan eleven bänka minst hundra kilo, samt ha bra bollkontroll (vara bra på fotboll)

På förekommen anledning: betygskriterier för betyget A i idrott och hälsa på gymnasiet: Eleven kan med säkerhet och med goda rörelsekvaliteter genomföra en bredd av aktiviteter, även av komplex karaktär, som utvecklar den kroppsliga förmågan

@AmandaAskell Which countries would be good examples of this?

Extremely racially homogeneous countries that think of themselves as bastions of tolerance are often one wave of immigrants away from finding out that their tolerance wasn't necessarily strong, it was just completely untested.

@johbirgmeier @emollick I disagree. The development in recent years has been very rapid, and if one simply extrapolates that this rate of development continues, plus the enormous sums of money that are now being spent on developing better models, his forecast does not seem unreasonable.

When the CEO of one of the most advanced AI companies in the world is telling us that AI will reach the ability levels of educated humans in 2-3 years, we should update our assumptions about the near future of AI It doesn't mean that he is right, but we should take it seriously

@OHaggstrom Din föreläsning i Umeå var mycket bra. Gissar att denna är snarlik men ska kika ändå. :)

@RokoMijic @RichardMCNgo I believe he meant that having confidence can alter your IQ test score not you "actual" intelligence.

@RichardMCNgo Why would you score better than 100 just by being confident? I don't think IQ works like that! Confidence and persistence shape outcomes a lot but afaik they do not actually change your IQ

A reason to be cautious about IQ: it's very hard to disentangle from motivation. Imagine having IQ 100 in a world where only 1 in 10,000 had an IQ that high. You'd top all your classes, develop a lot of confidence, and score way better on tests than IQ 100 people in our society.

@JeffLadish @sidereal111 How does Claude+ compare to ChatGPT-4?

@sidereal111 I'm talking about ChatGPT 3.5 Which to be fair I haven't used that much since I mostly use GPT-4 or Claude But it seems in a similar reference class to 3.5

Hugging Chat, a ChatGPT-clone based on a LLaMA-based model, was just launched. I've been using it and while it's a little rough around the edges, it feels similar to ChatGPT in terms of capabilities Only 5 months passed between the launch of ChatGPT and HuggingChat

Fredagkvällen är räddad! Nu finns min och @eriklindqvist1:s artikel i Ekonomisk Debatt om lotterivinnare att läsa alldeles gratis. nationalekonomi.se/sites/default/…

@Varjo_Haukka @UniverseIce A matter of taste. I like MIUI much more than OneUI.

@tailcalled Out of curiosity, what are your main concerns?

@ReBoot23 @TechTabletscom @Tradingshenzhen Interesting. Have you seen any photo comparisons between 13 Pro and 12sU?

@ConnollyFors @TechTabletscom @Tradingshenzhen Many Chinese reviewers reported that Mi 11 Pro's GN2 tuning wasn't as good as the 11 Ultra.

It seems a lot of people forgot the Xiaomi 13 Pro is pretty much the Xiaomi 12S Ultra with a new look and S8Gen2 #Xiaomi13Series So if you want to save almost 500 euros and don't mind the Chinese version it's one way to go to get the 1-inch sensor. @Tradingshenzhen

@TechTabletscom @ReBoot23 @Tradingshenzhen Really? I had the Mi 11 Pro, which shares the same main sensor as the Mi 11 Ultra, and the 12sU simultaneously. To my surprise, the photos taken with the main sensor on the 12sU were significantly sharper and more visually appealing.

@ReBoot23 @Tradingshenzhen Not really I've had both. The best for cameras from them is the Mi 11 Ultra

@tailcalled It appears that I was mistaken. After conducting a brief search on Google Scholar, it seems that agreeableness may not be as strongly associated with political ideology as I had previously thought.

@tailcalled Low empathy/ agreeableness will increase the likelihood of right-wing sympathies.

@f_hieronymus Bidragsmiljonen känns intuitivt betydligt mer fel. Måste fundera på om det är rimligt...

Vad är värst? 1.) En privatperson undanhåller en miljon i skatteintäkter från staten och använder det för privat konsumtion. 2.) En privatperson lurar till sig en miljon i bidrag från det offentliga och använder den till privat konsumtion.

@richardohrvall Absolut

@ConnollyFors Intressant, du får gärna säga till när ni har några resultat.

Förra veckan släpptes data från ESS som samlades in 2020, vilket brukar vara en högtidsstund. För vissa länder, däribland Sverige, ändrades dock datainsamlingsmetoden från besöksintervju till enkät. Det är därför lite svårt att veta hur man ska tolka förändringar.

@richardohrvall Vi jobbar på att försöka undersöka vad ev. metodeffekter beror på. En hypotes är att skillnader beror på selektionseffekter (högtillitare mer benägna att ställa upp på face to face), en annan att det beror på social önskvärdhet (starkare vid face to face).

@richardohrvall Tror inte det. Fick den informationen på ett projektmöte. Metodbytet i Sverige ansågs nödvändigt givet covid-19 plus den långa trenden med kraftigt fallande svarsfrekvens för face-to-face.