Dejan Dular (@dejandular) profile picture

May the farce be with you!

Slovenia
Joined June 2009
Dejan Dular retweeted
This is truly interesting!
Chomsky was wrong. Researchers from UC Berkeley and Google DeepMind cracked one of the biggest mysteries in AI. For years, scientists have been finding impossible geometric shapes hidden inside the brains of LLMs. Calendar months arrange themselves into literal circles. Historical years bend into smooth, rippled timelines. Cities organize themselves into actual 2D maps of latitude and longitude. Everyone assumed this proved massive transformers were spontaneously inventing complex "world models" out of thin air. They were wrong. This paper proves you don't need a massive LLM for any of this to happen. Even basic, decade-old word embeddings like word2vec produce the exact same circles, timelines, and maps. The geometry isn't coming from the AI's architecture. It’s coming from a hidden mathematical property inside human language itself. Researchers call it translation symmetry. In natural text, how often any two months appear near each other depends almost entirely on the time gap between them. January and February share the same statistical relationship as July and August. When a neural network trains on data with that kind of symmetry, the math forces it to learn Fourier representations, literal sine and cosine waves. The circles? Those are the long waves. The strange "ripples" researchers found on AI timelines? Just higher harmonics. Here is the craziest part. Rresearchers tried to break it. They deleted every single sentence in the training data where two months appeared together. No "January" next to "February." No "March" next to "April." And the circle formed anyway. Why? Because the geometry is a collective phenomenon across the entire language. Even if two months never touch in a sentence, they are both pulled into orbit by thousands of other words tied to the same seasons, weather, and events. The rest of human language forces the shape into existence.
1
1
1
206
Kill me naou.
2
2
182
Lisice zdaj hodijo naokoli s cvikcangami.
1
40
FYI: Ne kupovat Smeg naprav iz tujine, ker ni uradnega servisa pri nas. Servis v garancijskem roku tako postane nočna mora. Pri SLO uvozniku (Steel Plast) so super prijazni, ampak niso uradni serviser in ne izvajajo garancijskih servisov za izdelke, ki niso kupljeni pri njih.
1
41
Pomoč! Katero podjetje bi znalo obnoviti fotografije iz android telefona, ki je enostavno rečeno - umrl? Neki poskusi obnovitve so bili, vendar neuspešni.
2
2
383
Kdo ve kaj se je zgodilo z aplikacijo Dež? Na Google Play Store ni več na voljo.
2
1
137
I used BETWEEN in my SQL query and I am not even ashamed!
70
Today's #AdventOfCode (Day 11) is an excellent example of why one should not just write the code ad hoc. At first, it seemed that the code would take hours(days) and consume all resources. Then using a lateral approach, everything was done in 20ms.
2
97
All about the sensitive data by @evdlaar #DataSatSlo
1
6
209
Mastering Data Cleaning in Fabric by @KarolPapaj #DataSatSlo
2
4
128
Today, the chef of the house is @spaghettidba with Baking Extended Events recipes #DataSatSlo
1
5
123
Automating Power BI with PowerShell @FrankGeisler #datasatslo – at Republic of Slovenia
1
3
6
251
And another one from @FrankGeisler today. Now he is introducing you to Fabric Real-Time Intelligence. #DataSatSlo – at Republic of Slovenia
3
5
154
Fabric Real-Time ainetlligence for Pober BI pros discussed by @SQLMissSunshine and Devang Shah – at Republic of Slovenia
1
3
122
Do the commits the correct way by @sqldbawithbeard #DataSatSlo
2
3
121
Uwe Ricken describing How to properly handle LOB data in SQL server. #DataSatSlo – at Republic of Slovenia
2
4
143
Full room at #DataSatSlo @SasaPopovicMSFT and @Strale15 taking you from SQL to Insights.
1
3
1
5
272
Pointing out the obvious @Kjonge speaking about Fabric security. #datasatslo
2
5
144