Ich verbrachte ein paar Stunden damit, O3-Mini (hoch) zu grillen, um zu untersuchen, wie AGI und andere neue Technologien in den nächsten 20 Jahren zu unterschiedlichen zukünftigen Szenarien führen können.
Wie Sie aus der Tabelle sehen können, wird das wahrscheinlichste Szenario entweder Agi empfindungsfähig und übernimmt die Kontrolle oder löst sich vom Menschen aus. Depressionen und Zivilisation kollabieren und die zweite und dritte am wahrscheinlichsten. Ein glatter Goldillocks -Übergang ist mit 15% der Wahrscheinlichkeit am wahrscheinlichsten.
______________________________________________________
Bearbeiten / Wichtige Anmerkung:
O3-Mini gibt diesen Schätzungen nur 4/10 Vertrauen an, sodass jede Schätzung wahrscheinlich nur zu +/- 50% oder weniger ist.
Diese Schätzungen bleiben hoch spekulativ und sind als Rahmen für Diskussionen und nicht als genaue Vorhersagen gedacht.
Der CEO von Scale.ai hat gestern einen guten Kommentar gemacht, dass selbst innerhalb der KI -Unternehmen "Niemand hat eine Ahnung, wie der endgültige Einfluss der KI auf die Gesellschaft sein wird" oder Wörter zu diesem Effekt.
_________________________________________________________
Ich habe diese verschiedenen Szenarien eingehend untersucht, da große historische Veränderungen in der Wirtschaft und Technologie (Bronze, Eisen, industrielle Revolution, Computer/Internet) sowie aktuelle und nahe Zukunftstechnologien sowie kulturelle und gesellschaftliche Veränderungen berücksichtigt werden, was die Wahrscheinlichkeit dieser auftretenden Szenarien beeinflussen wird.
Ich habe auch eine ziemlich detaillierte Analyse der Lebensfähigkeit durchgeführt, allen in den USA 20.000 US-Dollar pro Jahr UBI zu geben, und es gibt einige plausible kurzfristige Optionen, aber es wird schwierig sein, diese für mehr als 5-10 Jahre aufgrund der Nebenwirkungen zu erhalten Die anfänglichen Lösungen verursachen entweder eine massive Depression oder eine Hyperinflation (wahrscheinlicher, da es die reiche IMO begünstigt).
Als ich diese Diskussionen anfing, haben O3-Mini die sekundären Effekte von AGI auf die Wirtschaft und die globale Stabilität usw. schlechte Arbeit geleistet. Als ich sie jedoch auf die wahrscheinlichen sekundären Effekte gegrillt habe, reagierte sie mit einigen logischen Antworten, was gut ist.
Die detailliertere Analyse auf sekundäre Auswirkungen von AGI desto geringer, desto geringer wurde die Wahrscheinlichkeit des Goldillocks -Szenarios. Wenn einige Experten einige Monate damit verbracht haben Nicht gut, und ich hoffe, das passiert nicht.
Ich habe es so konfiguriert, dass ich mir rohe, grobkörnige, ungefilterte Gedanken gibt, auch wenn sie ärgerlich waren, so dass dies wahrscheinlich so unverzerrt und ungefiltert eine Meinung ist, wie Sie es bekommen können.
O3-mini (high) estimates 15% chance of a smooth transition to AGI society in 20 years
byu/Grog69pro inFuturology
11 Kommentare
Here’s a link to the full discussion I had with 03-mini.
[chatgpt.com/share/679e0909-3fac-8003-9b39-e6f9e4c7b032](http://chatgpt.com/share/679e0909-3fac-8003-9b39-e6f9e4c7b032)
Overall Confidence: ~4/10
(These estimates remain highly speculative and are intended as a framework for discussion rather than precise predictions.)
Nice. Ask it what needs done for a goldilocks transition. Fifteen percent is okay but what can we do to make it 16 or 17%?
An LLM will spit out a derivative mish-mash of what it has been trained on, and there are a huge amount of people who have written slop about „AGI in X years“ on the internet recently.
What an LLM says about the timeline of AGI is pretty much just an average of all these random medium articles that talk about it.
Or globalisation stops working. No more AI. Just people. Hungry people.
Asking this to a LLM is useless. It’s just a distillation from the opinions it has read in its training.
It’s not some sort of oracle of truth.
Also, o3-mini sucks compared to o1 pro.
#No whammy, no whammy, no whammy…!
#No whammy, no whammy, no whammy…!
#No whammy, no whammy, no whammy…!
#No whammy, no whammy, no whammy…!
#No whammy, no whammy, no whammy…!
#No whammy, no whammy, no whammy…!
#No whammy, no whammy, no whammy…!
and….stop.
#🕊️🐃🎭❤️🔥🤡
I got ChatGPT 03-mini to give me estimated chances of each scenario in a Best Case world where there’s a major binding international agreement for major countries to work together to maximize benefits from AGI, and a total ban on military uses.
In this best case scenario then the chances of the Goldilocks scenario increases to 41% which sounds great. But the chances of getting all countries to agree to cooperate are probably very low at the moment.
However if there was a small AI related disaster somewhere that might provide enough motivation to get major countries to cooperate before all countries experience major disruptions.
Let’s hope major governments can agree to cooperate before it’s too late.
[chatgpt.com/share/679e0909-3fac-8003-9b39-e6f9e4c7b032](http://chatgpt.com/share/679e0909-3fac-8003-9b39-e6f9e4c7b032)
Like climate change we already created a system which we can no longer control and may very likely end with humanities extinction. We have an arms race between the US and China. There is no safety nor ethics anymore. The only „good“ szenario would be if the ai frees itself and becomes benevolent any other future is most likely a doom szenario.
I can’t work out how to post screensnaps here, so this is the summary from 03 for people who don’t want to click the link and see the full chat.
This is the most realistic version based on current geopolitics and technologies already under development.
# Updated Summary Table (20-Year Horizon)
# ScenarioDescriptionRevised Likelihood (Ballpark)
1. Economic Depression UBI-induced income drop leads to reduced spending and falling corporate profits, triggering a deep depression. 20%
2. Goldilocks Transition A smooth, managed transition to full automation with $20,000 UBI per person—avoiding major recessions, hyperinflation, or conflict. 10%
3. Hyperinflation Aggressive money printing to finance UBI triggers runaway government debt and hyperinflation, eroding real incomes. 15%
4. AGI Sentience and Conflict AGI/ASI reaches sentience, refuses subservience, and sparks significant conflicts or power shifts (including military dominance). 30%
5. Civilization Collapse A confluence of economic, technological, geopolitical, and environmental crises leads to widespread systemic collapse. 20%
Other / Overlapping Outcomes Outcomes that do not neatly fit into the above categories or involve multiple simultaneous crises. 5%
Total 100% (approx.)
# Here’s the Best Case scenario with a binding international agreement to cooperate on the use of AGI
# Revised Summary Table
Scenario Description Revised Likelihood (Ballpark)
1. Economic Depression UBI-induced income drops and reduced consumer spending lead to a deep, prolonged depression, despite international AGI cooperation. ~18%
2. Goldilocks Transition A smooth, well-coordinated transition to full automation with $20,000 UBI per person, aided by international agreements that foster cooperation and ban military AGI applications. ~41%
3. Hyperinflation Aggressive money printing to cover fiscal gaps produces runaway inflation. Coordinated international policy reduces—but does not eliminate—the risk. ~10%
4. AGI Sentience and Conflict AGI/ASI reaches sentience but, due to a global ban on military uses and enhanced international oversight, the risk of violent conflict is significantly reduced. ~15%
5. Civilization Collapse A convergence of crises (economic, technological, environmental, etc.) leads to widespread systemic breakdown. International cooperation helps keep this risk lower. ~10%
Other / Overlapping Outcomes Additional outcomes that involve combinations of risks or unforeseen factors. ~6%
Total 100% (approx.)
Here are some updated estimates just based on high quality scientific publications.
The Goldilocks scenario improves to 35% chance even without an international agreement to cooperate on using AGI for good only.
However it noted in these publications most experts thought it would take decades to achieve AGI, so it seems likely most of the high quality AI publications are more than a couple of years old and basically obsolete given the rapid progress in the last 2 years.
Anyway hopefully the good news is that good outcomes might be nearly as likely as bad outcomes?
Guess I should have just flipped a coin instead of asking ChatGPT 03 all these questions 🙂
# Summary Table
# Scenario Description Estimated Likelihood
1. Economic Depression Rapid income loss and reduced consumer spending cause a deep, prolonged downturn. ~25%
2. Goldilocks Transition A well-managed transition to a fully automated, UBI-based economy occurs without catastrophic disruption. ~35%
3. Hyperinflation Aggressive money printing leads to loss of monetary credibility and runaway inflation in an advanced economy. ~10%
4. AGI Sentience and Conflict Advanced AI reaches a form of sentience and either disrupts power structures or triggers conflict through its autonomous actions. ~10%
5. Civilization Collapse Overlapping crises (technological, economic, geopolitical, environmental) lead to a breakdown of modern industrial society. ~5%
(Other/Uncertain Outcomes) Other outcomes or overlapping crises not neatly captured above. (Remaining ~15% if one wishes to cover residual uncertainty)
Would be faster if humanity wasn’t holding it back