Just in time for our final review of the year, we at the agency are taking another look back at what\u2019s new so far. One of the topics we\u2019ve been discus\u00adsing is ChatGTP and the new Thin\u00adking Mode. Could it make ever\u00ady\u00adthing better? Is there more to AI than suggesting shor\u00adtenings and checking for spel\u00adling and grammar mistakes? Could our virtual colle\u00adague even write the basis for an entire profes\u00adsional article\u2014without any hallucinations?<\/p>\n
In prac\u00adtice, it looks like this:<\/b><\/p>\n
The test run starts off well thought out. We write a detailed prompt, define the target audi\u00adence, topic, tone, and length. Of course, we put a lot of effort into this, because we wanted to be able to continue working with the results.<\/p>\n
Then we wait. Until now, neither the state\u00adments nor the sources provided by the AI could be trusted. Our eager colle\u00adague ChatGTP likes to make things up to make herself look better. This shouldn\u2019t happen anymore in the Thin\u00adking mode. We are excited to see the result. But then what happens is<\/p>\n
Silence. Abort. End.<\/b><\/p>\n
Either the text breaks off in the middle or, even better, nothing<\/b> appears\u2014Thinking Mode thinks and thinks, and in the end, there is nothing. No sentence, no frag\u00adment. And even the sentence \u201cPlease finish writing this text\u201d doesn\u2019t work. Because the begin\u00adning has long since vanished into virtual fog.<\/p>\n
Why the thin\u00adking mode some\u00adtimes \u201cstrays\u201d<\/b><\/p>\n
In theory, the thin\u00adking mode is a bles\u00adsing, because it actually follows instruc\u00adtions such as \u201cdon\u2019t esti\u00admate the word count, just count\u201d or \u201cdon\u2019t inter\u00adpret the results, just repro\u00adduce what is actually there.\u201d The problem: addi\u00adtional \u201cthin\u00adking work\u201d requires resources. It costs compu\u00adting time, and on top of that, each answer has a limited \u201cword budget\u201d (number of tokens). If this budget is exceeded or the calcu\u00adla\u00adtion takes too long, it\u2019s game over \u2013 the tech\u00adno\u00adlogy says, \u201cTime\u2019s up.\u201d<\/p>\n
For us, this feels like a down\u00adload that stops at 98%\u2014painful for our time budget, emotio\u00adnally\u2026 let\u2019s say: not good for our blood pressure.<\/p>\n
Avoi\u00adding blackout experiences<\/b><\/p>\n
At the end of the day, we\u2019ve at least made some progress with this:<\/p>\n
ChatGPT is not an author, but a struc\u00adtural aid, a wording machine, or a source of ideas for varia\u00adtions, subhea\u00addings, and intro\u00adduc\u00adtions. What it defi\u00adni\u00adtely does not replace: a know\u00adled\u00adgeable author. In any case, we prefer to let AI continue to do the smaller auxi\u00adliary tasks \u2013 that saves time and doesn\u2019t waste it.<\/p>\n","protected":false},"excerpt":{"rendered":"
Just in time for our final review of the year, we at the agency are taking another look back at what\u2019s new so far. One of the topics we\u2019ve been discus\u00adsing is ChatGTP and the new Thin\u00adking Mode. Could it make ever\u00ady\u00adthing better? Is there more to AI than suggesting shor\u00adtenings and checking for spelling [\u2026]<\/p>\n","protected":false},"author":4,"featured_media":1938,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_et_pb_use_builder":"","_et_pb_old_content":"","_et_gb_content_width":"","wp_typography_post_enhancements_disabled":false,"footnotes":""},"categories":[20,11],"tags":[],"class_list":["post-1936","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-blog-en","category-uncategorized"],"yoast_head":"\n