22.09.2026
Setzen Sie sich in 4er Gruppen zusammen.
Texte werden in 0 oder 1 gespeichert. Mit welchem vordefinierten Schema können Sie einen beliebigen Text so effizient wie möglich speichern?
| Character | Unicode code point | UTF-8 hex | UTF-8 binary |
|---|---|---|---|
| A | U+0041 | 41 | 01000001 |
| M | U+004D | 4D | 01001101 |
| Z | U+005A | 5A | 01011010 |
| a | U+0061 | 61 | 01100001 |
| m | U+006D | 6D | 01101101 |
| z | U+007A | 7A | 01111010 |
|
U+0020 | 20 | 00100000 |
\n |
U+000A | 0A | 00001010 |
from transformers import AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained("openai-community/gpt2")
texts = [
"A",
"cat",
"Hello world",
" Hello world",
"believable",
"unbelievable",
"a\nb"
]
for text in texts:
ids = tokenizer.encode(text, add_special_tokens=False)
tokens = tokenizer.convert_ids_to_tokens(ids)
print("Text:", repr(text))
print("Tokens:", tokens)
print("Token IDs:", ids)
print()
Text: 'A'
Tokens: ['A']
Token IDs: [32]
Text: 'cat'
Tokens: ['cat']
Token IDs: [9246]
Text: 'Hello world'
Tokens: ['Hello', 'Ġworld']
Token IDs: [15496, 995]
Text: ' Hello world'
Tokens: ['ĠHello', 'Ġworld']
Token IDs: [18435, 995]
Text: 'believable'
Tokens: ['bel', 'iev', 'able']
Token IDs: [6667, 11203, 540]
Text: 'unbelievable'
Tokens: ['un', 'bel', 'iev', 'able']
Token IDs: [403, 6667, 11203, 540]
Text: 'a\nb'
Tokens: ['a', 'Ċ', 'b']
Token IDs: [64, 198, 65]
GPT 2: 50’257 tokens
Modern: > 1’000’000 tokens
https://platform.openai.com/tokenizer
Erstellen Sie einen Satz, geben Sie dem Nachbar die ersten paar Tokens und lassen Sie ihn den nächsten Token erraten.
from transformers import AutoTokenizer, AutoModel
import torch
tokenizer = AutoTokenizer.from_pretrained("openai-community/gpt2")
model = AutoModel.from_pretrained("openai-community/gpt2")
model.eval()
texts = [
"A",
"cat"
]
embedding_layer = model.get_input_embeddings()
for text in texts:
inputs = tokenizer(text, return_tensors="pt", add_special_tokens=False)
ids = inputs["input_ids"][0]
tokens = tokenizer.convert_ids_to_tokens(ids)
with torch.no_grad():
# Raw input embeddings: directly from the embedding table
embeddings = embedding_layer(ids)
print("Text:", repr(text))
print("Tokens:", tokens)
print("Token IDs:", ids.tolist())
print("Embedding shape:", embeddings.shape)
print("The vector is:")
print(embeddings[0].tolist())
print()Text: 'cat'
Tokens: ['cat']
Token IDs: [9246]
Embedding shape: torch.Size([1, 768])
The vector is:
[-0.016361085698008537, -0.09340496361255646, 0.2424590289592743, 0.1397625356912613, 0.03876500949263573, -0.2591763436794281, -0.27238377928733826, -0.16245122253894806, 0.16829444468021393, 0.08293303847312927, 0.013625058345496655, -0.27883821725845337, 0.14928586781024933, 0.14083561301231384, 0.05570671334862709, -0.3691481649875641, 0.22002798318862915, -0.042841047048568726, 0.2205713540315628, 0.08649542182683945, 0.12367253750562668, -0.1499468833208084, 0.14462877810001373, -0.11504415422677994, -0.14248539507389069, -0.07152555882930756, -0.0525745190680027, 0.15497425198554993, -0.06780064851045609, -0.2058633267879486, 0.20652440190315247, -0.02965833991765976, 0.08339905738830566, -0.04833613336086273, 0.12067288160324097, 0.1974879652261734, -0.3193340003490448, 0.012369612231850624, 0.10672488063573837, -0.04727446660399437, -0.3036808371543884, 0.11387067288160324, 0.09488601237535477, -0.21753600239753723, 0.07961409538984299, -0.09413309395313263, -0.03938238322734833, -0.07039275020360947, 0.20333722233772278, -0.1554637849330902, 0.2927825152873993, -0.07704915851354599, 0.07868365198373795, 0.12144948542118073, 0.15276168286800385, -0.14643165469169617, 0.4247491657733917, 0.19212593138217926, -0.041541408747434616, -0.0850478857755661, -0.2787071466445923, 0.06556132435798645, -0.20258086919784546, 0.18557333946228027, 0.13530132174491882, -0.08202585577964783, -0.06389398127794266, 0.07009650766849518, 0.16802138090133667, 0.05966462194919586, 0.3264687657356262, -0.10995844006538391, 0.10556178539991379, 0.1844596564769745, -0.11555470526218414, 0.005354898516088724, 0.06634213775396347, 0.18418389558792114, -0.1068630963563919, 0.04913080856204033, -0.08526840806007385, -0.2518639862537384, 0.003143342910334468, 0.1804843693971634, 0.15052711963653564, 0.04419228062033653, -0.2426886111497879, 0.11044622212648392, 0.0970262810587883, 0.11234410107135773, -0.1518537700176239, -0.14437256753444672, 0.23230873048305511, -0.024112675338983536, -0.06765676289796829, 0.1156836450099945, -0.266793429851532, -0.12285173684358597, 0.11203598231077194, 0.06005825847387314, -0.05346614494919777, 0.12588262557983398, -0.09662143886089325, -0.19754476845264435, -0.2031158208847046, 0.13232780992984772, 0.017606088891625404, -0.1331987977027893, 0.11589977890253067, 0.10372360050678253, 0.07220403850078583, 0.1644088327884674, -0.07746899127960205, -0.022746911272406578, 0.1145697683095932, 0.005957134999334812, 0.39594972133636475, -0.08284942060709, 0.012456733733415604, 0.0415172353386879, -0.014682512730360031, -0.135150745511055, -0.05785886570811272, 0.042313102632761, -0.0793331190943718, -0.2702028453350067, 0.2805854380130768, -0.07441112399101257, 0.11179929226636887, 0.09083987772464752, 0.06389196962118149, -0.08819812536239624, -0.01904943957924843, -0.13858816027641296, -0.04897405207157135, -0.17850333452224731, 0.14161550998687744, 0.04971975088119507, -0.04605376720428467, -0.15443900227546692, 0.06617309153079987, -0.05384760722517967, 0.09921273589134216, 0.13080994784832, -0.08854424953460693, -0.2840065062046051, -0.02972940169274807, -0.08819353580474854, -0.03396836295723915, -0.14952079951763153, 0.02950727380812168, 0.06999912858009338, 0.06613373756408691, -0.12824204564094543, -0.05458778887987137, -0.1391710489988327, -0.13682205975055695, 0.03527950122952461, 0.08139467984437943, -0.14360448718070984, -0.05591137707233429, 0.15230049192905426, -0.07797905802726746, 0.2561819553375244, 0.016448795795440674, -0.04333009943366051, 0.04680568352341652, 0.2895946204662323, 0.006858479231595993, 0.2135537564754486, 0.03780953586101532, -0.162540465593338, -0.04209734871983528, -0.010930346325039864, -0.038604531437158585, 0.04526738449931145, 0.2572128474712372, 0.032335713505744934, -0.12058413773775101, 0.01354928221553564, -0.01707439497113228, 0.04041294381022453, -0.2544141113758087, 0.14550462365150452, 0.32653355598449707, -0.05447030067443848, 0.08872339874505997, -0.032053422182798386, 0.1484597772359848, -0.06985272467136383, -0.0605902224779129, 0.21767176687717438, -0.15659214556217194, -0.0619233101606369, -0.2654725909233093, 0.24714842438697815, -0.12134813517332077, -0.07410012185573578, 0.1074477881193161, -0.024348974227905273, -0.2747279703617096, 0.18277612328529358, 0.004594720900058746, 0.14255641400814056, 0.020123416557908058, 0.04133181273937225, -0.018891766667366028, -0.0070403022691607475, -0.03381647914648056, -0.13835126161575317, 0.026912670582532883, 0.14468277990818024, -0.021608944982290268, -0.004169708117842674, -0.03634539991617203, -0.057931527495384216, -0.0908757820725441, -0.13588868081569672, 0.14068809151649475, 0.14214162528514862, 0.004056630190461874, 0.01000190619379282, 0.09966544806957245, -0.0718214139342308, -0.09584313631057739, 0.005125546362251043, -0.2575571537017822, 0.19795745611190796, 0.1544952690601349, 0.022637637332081795, -0.25213515758514404, -0.10910143703222275, -0.14668214321136475, -0.11400547623634338, 0.1382637321949005, -0.19515198469161987, 0.055449772626161575, -0.0036417068913578987, -0.29149243235588074, -0.164504736661911, 0.046878427267074585, -0.225123792886734, 0.20002727210521698, 0.10697862505912781, 0.16506700217723846, -0.07809083163738251, 0.15108534693717957, -0.009537586942315102, 0.09249945729970932, 0.07756288349628448, 0.16313457489013672, 0.15626922249794006, 0.028588488698005676, -0.11567679047584534, 0.03491910174489021, 0.0033353520557284355, -0.08695478737354279, -0.08638929575681686, 0.1233266070485115, -0.0691155344247818, -0.045821744948625565, -0.06007238104939461, 0.050143346190452576, -0.1449822336435318, -0.24250608682632446, -0.07734102755784988, 0.11820480227470398, 0.1351100504398346, 0.19037167727947235, 0.17455287277698517, 0.09251242130994797, -0.12529748678207397, -0.11490848660469055, -0.13117901980876923, -0.21701258420944214, 0.06819556653499603, -0.012136389501392841, -0.17737486958503723, -0.08566293865442276, -0.1905902922153473, 0.2842336595058441, -0.04100750759243965, 0.05296124145388603, 0.047999344766139984, -0.06414822489023209, -0.09112175554037094, 0.29074209928512573, -0.2503187656402588, -0.1084781065583229, 0.17534638941287994, 0.06095774471759796, 0.0465896837413311, 0.009666230529546738, -0.1299930363893509, -0.027251724153757095, 0.049787331372499466, -0.06186866760253906, -0.18665863573551178, -0.07687986642122269, -0.10912930220365524, -0.04098173603415489, -0.06172481179237366, -0.053737953305244446, 0.05815371870994568, -0.09863217920064926, -0.26551467180252075, 0.12357717752456665, -0.002636916469782591, 0.0443752147257328, -0.10180223733186722, -0.16518911719322205, 0.017386866733431816, -0.25606292486190796, 0.043996330350637436, 0.20479539036750793, 0.004895505961030722, -0.021953323855996132, -0.10307517647743225, -0.13869129121303558, 0.04934246465563774, 0.20477673411369324, 0.047282785177230835, 0.16300159692764282, -0.01954091712832451, -0.07137879729270935, -0.07127375900745392, -0.2524056136608124, 0.08521950989961624, -0.16815339028835297, 0.07129006832838058, 0.130120187997818, -0.10878024995326996, -0.018849525600671768, 0.009248165413737297, 0.007848549634218216, 0.22126919031143188, 0.06376611441373825, -0.16165275871753693, -0.036549150943756104, -0.09231092035770416, -0.10518846660852432, 0.11083681881427765, -0.11745647341012955, -0.0016450758557766676, -0.02577287331223488, 0.09020651876926422, 0.10887619853019714, 0.16845135390758514, -0.2664169371128082, -0.030941037461161613, -0.01871739886701107, -0.06784279644489288, -0.142384335398674, -0.0026496881619095802, -0.06233548745512962, -0.0574648417532444, -0.10092271119356155, 0.014226469211280346, -0.1949736773967743, 0.008493809029459953, -0.14015188813209534, 0.03712889924645424, -0.4071781039237976, -0.0478425994515419, 0.4013194143772125, 0.3211887776851654, 0.10507680475711823, 0.03491642326116562, -0.1302030384540558, -0.029799172654747963, -0.17383332550525665, 0.06918296962976456, -0.2638380229473114, 0.1268089860677719, 0.17730343341827393, -0.109433613717556, 0.07369467616081238, -0.04598681256175041, 0.18701305985450745, -0.060537099838256836, -0.13076302409172058, -0.09196111559867859, -0.02904343791306019, -0.05416225269436836, 0.12136288732290268, -0.03084559366106987, -0.11734398454427719, -0.21271629631519318, 0.020885422825813293, 0.2910820245742798, -0.17514750361442566, 0.046861160546541214, 0.07395925372838974, -0.13225147128105164, 0.028280222788453102, 0.21246756613254547, 0.18701842427253723, 0.09781043976545334, 0.17992094159126282, 0.26694661378860474, 0.17093543708324432, -0.11910172551870346, -0.20219770073890686, 0.044475287199020386, 0.06009715422987938, -0.18201203644275665, 0.022370221093297005, 0.19018450379371643, -0.3198891878128052, -0.2551420032978058, 0.07953917235136032, 0.08136606216430664, 0.12448912858963013, 0.08713237196207047, -0.04553601145744324, -0.23420341312885284, 0.11666262894868851, 0.08702189475297928, 0.025680648162961006, -0.20734798908233643, 0.18486300110816956, -0.018431687727570534, 0.049841225147247314, -0.1422727257013321, -0.06821028888225555, 0.13858897984027863, 0.04061681032180786, -0.03245803713798523, 0.21790796518325806, -0.056724026799201965, -0.25683629512786865, -0.09347919374704361, -0.04531891644001007, -0.1317279189825058, 0.068155437707901, -0.27211079001426697, -0.2026086449623108, -0.05646788701415062, -0.013355241157114506, -0.04228202626109123, -0.04154609516263008, -0.055999014526605606, -0.1521988809108734, 0.16170743107795715, -0.07531152665615082, -0.19667938351631165, 0.053582578897476196, -0.09884798526763916, 0.05389460176229477, -0.04893943667411804, 0.02958505228161812, 0.1271725296974182, 0.15667696297168732, 0.018528390675783157, -0.08550908416509628, -0.2336433380842209, 0.18592824041843414, 0.15278755128383636, -0.18235425651073456, -0.08338190615177155, -0.14144834876060486, -0.052624352276325226, 0.1744244396686554, -0.02897842787206173, -0.07534624636173248, -0.009970501996576786, -0.1702166199684143, -0.06763046979904175, 0.08557812869548798, 0.04925626143813133, -0.12563832104206085, -0.1651638001203537, -0.13169142603874207, 0.0677306279540062, 0.020861729979515076, -0.03456887975335121, 0.004781287629157305, 0.12093725055456161, 0.1958766132593155, 0.15197370946407318, 0.07925860583782196, -0.14918960630893707, 0.3141046166419983, 0.1525752693414688, -0.17319747805595398, -0.09137152880430222, 0.13390062749385834, -0.14101886749267578, -0.059530600905418396, -0.025033889338374138, -0.1136339008808136, -0.1206163614988327, -0.112601637840271, -0.04700680449604988, 0.18976610898971558, 0.05645706132054329, -0.20583580434322357, 0.03894269838929176, 0.01765262894332409, -0.2717885375022888, 0.20207126438617706, -0.077913798391819, 0.14437717199325562, 0.10467476397752762, -0.20957966148853302, -0.02104145474731922, 0.17907707393169403, -0.4005124270915985, -0.19312529265880585, 0.10826538503170013, 0.24648120999336243, 0.10261765867471695, -0.05029426887631416, 0.10469899326562881, 0.029899688437581062, -0.10433568805456161, 0.09644400328397751, 0.08517158031463623, -0.206651970744133, 0.12626570463180542, 0.206422358751297, 0.22478507459163666, 0.2739073634147644, -0.18809781968593597, -0.07453365623950958, 0.07694312185049057, 0.2994121015071869, 0.28032568097114563, 0.006251325365155935, 0.25848692655563354, -0.01760149374604225, 0.23183578252792358, -0.04323691129684448, 0.18887682259082794, -0.07658647745847702, 0.07513964921236038, -0.015742845833301544, 0.0517050139605999, 0.1274394690990448, -0.22347670793533325, -0.045028265565633774, 0.16056886315345764, 0.08764401078224182, 0.12397979199886322, 0.4416997730731964, -0.062466226518154144, 0.05908837541937828, -0.018098803237080574, 0.19964583218097687, 0.09586238861083984, -0.26234763860702515, -0.2825523316860199, 0.0022820779122412205, 0.18350552022457123, 0.19312357902526855, -0.10535218566656113, 0.18156012892723083, -0.15994378924369812, -0.08708500117063522, 0.011543862521648407, 0.23855069279670715, 0.016107626259326935, 0.057996444404125214, -0.0557699128985405, 0.09634870290756226, 0.12061449885368347, -0.34606459736824036, 0.0725790336728096, 0.03005097061395645, 0.10581577569246292, 0.05319081246852875, 0.05146980658173561, 0.02162906713783741, -0.05306810885667801, -0.02167229913175106, 0.05393480882048607, -0.01909218542277813, 0.06355657428503036, -0.15273496508598328, -0.16696777939796448, 0.07557927817106247, 0.016657520085573196, -0.0436907522380352, 0.00502001540735364, -0.1860574185848236, 0.03037017211318016, 0.2442237138748169, -0.012576671317219734, -0.2313520461320877, 0.156240314245224, 0.16345180571079254, -0.12061086297035217, -0.04280021786689758, 0.10787300020456314, 0.12155219912528992, -0.011280953884124756, 0.17571833729743958, -0.02351534366607666, 0.20489025115966797, -0.3030067980289459, 0.006661759223788977, -0.31568723917007446, 0.1435137242078781, -0.17374850809574127, -0.1698041707277298, 0.2276156097650528, -0.03598174825310707, 0.0047904024831950665, -0.2974128723144531, 0.202104389667511, 0.13797812163829803, 0.11289305984973907, -0.06263086199760437, 0.13471516966819763, 0.07288800179958344, 0.04810135439038277, -0.13974161446094513, 0.01967836357653141, -0.09318128973245621, 0.1716657429933548, -0.15187618136405945, -0.0554359145462513, -0.03442739322781563, 0.020096033811569214, 0.1315523087978363, 0.07433128356933594, -0.11893857270479202, 0.2786555588245392, 0.05967449024319649, -0.2073172628879547, -0.35549911856651306, -0.06447657197713852, -0.1325817108154297, 0.10935134440660477, 0.151214599609375, 0.024139029905200005, 0.06075652316212654, 0.03337767720222473, 0.13396388292312622, -0.050996121019124985, -0.01968289166688919, 0.0681430846452713, -0.1494400054216385, 0.24095620214939117, -0.10163936764001846, 0.11478090286254883, -0.12802991271018982, 0.05761439725756645, -0.01010966021567583, 0.19571518898010254, -0.28543218970298767, 0.023109126836061478, -0.028190160170197487, -0.010106373578310013, 0.038283511996269226, -0.008605421520769596, -0.11522215604782104, 0.1464390605688095, 0.03511372208595276, -0.21893106400966644, 0.2156209945678711, 0.07224100828170776, 0.08813518285751343, 0.23599788546562195, 0.11816892772912979, 0.06762979924678802, 0.05060616880655289, -0.022593209519982338, -0.2842129170894623, -0.27805382013320923, 0.07910069823265076, 0.19577139616012573, 0.04700881242752075, -0.0645257979631424, 0.012892939150333405, 0.287771612405777, 0.13391640782356262, -0.2548004984855652, -0.08546338975429535, -0.1514623612165451, 0.14419230818748474, -0.16941767930984497, -0.28641778230667114, -0.1720186024904251, -0.0473054014146328, -0.13245688378810883, 0.23898404836654663, 0.07356103509664536, -0.19753938913345337, -0.11407563090324402, -0.1408894807100296, 0.19006043672561646, -0.0673261433839798, 0.08869053423404694, 0.21322613954544067, 0.18689054250717163, -0.2635023295879364, 0.11484581977128983, -0.009960507974028587, 0.0344177708029747, -0.1273479014635086, -0.0943353921175003, 0.16983115673065186, 0.1510799080133438, -0.22476091980934143, 0.049499206244945526, 0.1384216845035553, 0.1259833723306656, 0.07865370064973831, 0.3356500267982483, 0.2287849485874176, -0.08848802745342255, 0.06221172213554382, -0.2150619775056839, 0.05525168776512146, 0.06896256655454636, 0.25680992007255554, -0.26370730996131897, -0.04600958153605461, 0.09111302345991135, -0.021443216130137444, 0.17124398052692413, 0.034207914024591446, -0.043751828372478485, -0.14573682844638824, -0.09525427222251892, 0.028924770653247833, -0.08300846815109253, -0.059136051684617996, -0.024387149140238762, 0.12586353719234467, 0.007528270594775677, -0.10518844425678253, 0.11190776526927948, 0.062092553824186325, -0.23821991682052612, 0.17628256976604462, -0.11317327618598938, -0.12008192390203476, 0.016212722286581993, -0.18070438504219055, 0.1419370472431183, 0.07490269094705582, -0.231834277510643, 0.013248786330223083, 0.2532402276992798, -0.10278435051441193, 0.05564931035041809, -0.021376635879278183, 0.1973973512649536, -0.15729038417339325, 0.18605276942253113, -0.17291545867919922, -0.07370968908071518, -0.14987541735172272, -0.15329447388648987, 0.0010656218510121107, 0.06813380122184753, -0.18283240497112274, 0.05188002437353134, -0.0836610347032547, 0.0670880600810051, -0.18670614063739777, 0.0011862353421747684, 0.037689339369535446, 0.10606680065393448, -0.1713053286075592, -0.15785232186317444]
\[ \underbrace{\text{``The cat''}}_{\text{Satz}} \;\rightarrow\; \underbrace{[\text{``The''}, \text{`` cat''}]}_{\text{Tokens}} \;\rightarrow\; \underbrace{[791, 8415]}_{\text{Token-IDs}} \;\rightarrow\; \underbrace{[\mathbf{e}_{791}, \mathbf{e}_{8415}]}_{\text{Embeddings}} \]
\[ \mathbf{e}_{8415} = \begin{bmatrix} x_1 \\ x_2 \\ x_3 \\ \vdots \\ x_{768} \end{bmatrix} \in \mathbb{R}^{768} \quad \text{implemented as} \quad \texttt{float32}^{768} \]
\[ \underbrace{t_1,\; t_2,\; t_3}_{\text{too far back}} \quad \underbrace{t_4,\; t_5,\; t_6,\; t_7,\; t_8}_{\text{context window}} \quad \xrightarrow{ \begin{array}{c} \overbrace{ \mathbf{W} = \begin{bmatrix} w_{1,1} & \cdots \\ \vdots & \ddots \end{bmatrix} }^{\text{model weights}} \;\longrightarrow\; \overbrace{ \mathbf{A} = \begin{bmatrix} 0.10 \\ 0.25 \\ \vdots \\ 0.40 \end{bmatrix} }^{\text{attention weights}} \end{array} } \quad P(T_9 \mid t_4,\dots,t_8) \]
\[ \underbrace{\text{The cat sat on the}}_{\text{context}} \longrightarrow \begin{cases} \text{mat} & 0.42 \\ \text{chair} & 0.18 \\ \text{floor} & 0.15 \\ \text{table} & 0.10 \\ \text{other tokens} & 0.15 \end{cases} \xrightarrow{\text{softmax},\; T = 0.7} \quad \begin{cases} [\text{mat}] & 0.75 \\ [\text{chair}] & 0.10 \\ [\text{floor}] & 0.06 \\ [\text{table}] & 0.02 \\ [\text{other tokens}] & 0.01 \end{cases} \]
https://www.soekia.ch/GPT/
Geben Sie der Maschine die Aufforderung eine Geschichte zu einem Thema deiner Wahl zu schreiben.
Schauen Sie nach kurzer Zeit hinein und passen Sie die Auswahl an.

Supervised Fine Tuning
Menschen schreiben Fragen und optimale Antworten
Create Reward Model
Menschen beurteilen KI-Antworten
Reinforcement Learning
KI trainiert mit dem Belohnungsmodell
Bei anspruchsvolleren Aufgaben macht ein Prompt die Zwischenschritte sichtbar und überprüfbar.
Eine Sitzung beginnt um 08:15, dauert 90 Minuten und enthält 15 Minuten Pause. Wann endet sie?
Antwort: «Um 10:00 Uhr.»
Die Antwort kann stimmen – der Rechenweg bleibt jedoch unsichtbar.
Löse die Aufgabe schrittweise: Berechne zuerst die gesamte Dauer und addiere sie dann zur Startzeit.
Antwort: «90 + 15 = 105 Minuten. 08:15 + 1:45 = 10:00. Die Sitzung endet um 10:00 Uhr.»
Die Rechnung lässt sich unmittelbar prüfen.
Wann findet die nächste Sitzung statt?
Antwort: «Vermutlich am Dienstagmorgen.»
Plausibel, aber nicht überprüft.
ReAct = Reasoning + Acting
Abbildung 1: Vereinfachter Ablauf eines Agenten mit ReAct-Schleife und Arbeitsgedächtnis.
Abbildung 2: Vereinfachte Architektur eines Clawdbot-Agenten.
Ich brauche Freiwillige:
Wir vergleichen zum Schluss den Tokenverbrauch in CHF.
Anmerkungen:
Der Gesamtverbrauch ist gedeckelt auf 50CHF.
Die Tokens sind sofort nach dem Kurs nicht mehr gültig.
Abbildung 3: Anzahl Schüler:innen mit attestierter Diagnose. Gfs Bern (2026) (n = 8288).
Ziel: Wir machen uns schlau über die Graphik mit Hilfe von KI.
Software: ChatGPT / Codex (lokale App)
Modell: GPT 5.6
Abbildung 4: Anzahl Schüler:innen mit attestierter Diagnose. Gfs Bern (2026) (n = 8288).
Ziel: Wir binden benutzen Graphiken in wissenschaftlichen Dokumenten.
Software: ChatGPT / Codex (lokale App)
Modell: GPT 5.6