Hinweis auf die DSGVO: Auf unserer Seite werden keine Dritt-Anbieter-Cookies verwendet und nur Daten erfasst, welche für das Minimum an Board-Funktionalität notwendig sind.
Bevor Sie sich registrieren oder das Board verwenden, lesen Sie bitte zusätzlich die DSGVO-Erklärung, welche in der Navigationsleiste verlinkt ist.

Kurzfassung der unserer Meinung nach wichtigsten DSGVO-Punkte:
Es kann vorkommen, dass Benutzer eigenverantwortlich Videos oder sonstige Medien in ihren Beiträgen verlinken, welche beim Aufruf der Forenseite als Teil der Seite samt zugehörigem Material mitgeladen werden. Sollten Sie dies nicht wünschen, verwenden Sie beim Benutzen des Forums einen Blocker wie z.B. uMatrix, welcher das Laden von Inhaltsblöcken von Fremd-URLs effektiv unterbinden kann.
Wir blenden keine Werbung ein und schränken die Inhalte in keinster Weise bei Benutzung von Addblockern ein. Dadurch ist die Grundfunktionalität des Forums auch bei vollständigem Blockieren von Drittanbieter-Inhalten stets gegeben.

Cookies werden unsererseits nur verwendet um das Einloggen des Benutzers für die Dauer der Forenbenutzung zu speichern. Es steht dem Benutzer frei die Option 'Angemeldet bleiben' zu verwenden, damit der Cookie dauerhaft gespeichert bleibt und beim nächsten Besuch kein erneutes Einloggen mehr notwendig ist.
EMail-Adressen werden für Kontakt bei wichtigen Mitteilungen und zur Widerherstellung des Passwortes verwendet. Die verwendeten IPs können von uns ohne externe Hilfsmittel mit keiner realen Person in Verbindung gebracht werden und werden nach spätestens 7 Tagen gelöscht. Diese IPs werden höchstens verwendet um Neuanmeldungen unerwünschter oder gesperrter Nutzer zu identfizieren und zu unterbinden. Wir behalten uns daher vor bei Verdacht, die Frist für die IP-Löschung auf maximal 14 Tage zu verlängern.
Unsere Webseite läuft auf einem virtuellen Linux-Server, welcher von einem externen Anbieter gehostet wird. Etwaige Verstöße der DSGVO-Auflagen seitens dieses deutschen Hosters können wir nicht feststellen und somit auch nicht verfolgen.
Wir halten Backups unserer Datenbanken, welche in regelmäßigen Abständen als Schutz vor Katastrophen, Hackerangriffen und sonstigen Ausfällen erstellt werden. Sollte ein Nutzer die Löschung seiner Daten wünschen, betrachten wir es als Unzumutbar die Backups auch von den Daten zu befreien, da es sich hierbei um eine mehrtägiges Unterfangen handelt - dies ist für eine Einzelperson beim Betrieb eines privaten Forums nicht zumutbar möglich ohne das Backup komplett zu löschen.
Sollten Sie etwas gegen die dauerhafte anonyme Speicherung ihrer EMail-Adresse, ihres Pseudonyms und ihrer Beiträge in einem Backup haben, sehen Sie von der Registrierung in diesem Forum ab. Für Mitglieder, welche vor dem 25.05.2018 registriert waren steht jedoch das Recht im Raum, eine Löschung der Datenbank-Backups zu beantragen.



Wenn dies Ihr erster Besuch hier ist, lesen Sie bitte zunächst die FAQs sowie die wesentlichen Regeln zur Benutzung des Forums.
Um an den Diskussionen teilnehmen zu können, müssen Sie sich zunächst registrieren.

ChatGPT

Wissenschaftsgeschichte, Wissenschaftstheorie bzw. -philosophie, Technik
Antworten
Skeltek
Site Admin
Site Admin
Beiträge: 5081
Registriert: 25. Mär 2008, 23:51
Wohnort: Stuttgart, Germany
Kontaktdaten:

ChatGPT

Beitrag von Skeltek » 28. Feb 2023, 05:00

Hi,
wir haben glaube ich noch gar keinen Thread speziell hierzu. Hat jemand von euch schon irgendwelche interessanten Gespräche damit gehabt?
Gödel für Dummies:
  • Unentscheidbarkeit - Dieser Satz ist wahr.
  • Unvollständig - Aussage A: Es existiert nur ein Element A.
  • Widersprüchlich - Dieser Satz ist falsch.

Benutzeravatar
belgariath
Ehrenmitglied
Ehrenmitglied
Beiträge: 1076
Registriert: 11. Feb 2006, 11:40
Wohnort: Trappist 1e

Re: ChatGPT

Beitrag von belgariath » 28. Feb 2023, 12:02

Ich habe mir nur mal ein Interview mit ChatGPT durchgelesen.
Das war stellenweise richtig dämlich, zum Beispiel konnte die KI nicht die Regierungsbezirke von Bayern nennen, was eigentlich einfaches Wiki-Wissen ist.
Teilweise war die KI auch recht leichtgläubig bzw. wankelmütig. Wenn man ihr eine ihrer Behauptungen korrigiert sagt sie einfach sinngemäß ja du hast recht, das Gegenteil ist der Fall.
Stellenweise war sie aber auch gerade im Bereich Kreativität und jenseits des reinen Faktenwissens atemberaubend beeindruckend. Zum Beispiel hat die KI im Namen von Markus Söder einen Brief an das Christkind geschrieben, quasi über Wünsche und Sorgen des bayerischen Ministerpräsidenten. Der war richtig gut und glaubhaft, vom Ton her absolut treffend, auch inhaltlich im Stil von Söder.
Selber benutzt habe ich ChatGPT aber noch nicht.
Der harmonische Oszillator ist die Drosophila der Physiker (Carsten Honerkamp)
Eine Welle ist, was so wackelt (Andrei Pimenov)
Elektrodynamik ist ein Schlauch vieler Hamsterkäfige (Haye Hinrichsen)

Benutzeravatar
seeker
Ehrenadmin
Ehrenadmin
Beiträge: 8098
Registriert: 26. Dez 2009, 10:29

Re: ChatGPT

Beitrag von seeker » 28. Feb 2023, 12:47

Ich habe es selber noch nicht ausprobiert, muss ich noch machen.

Das hier fand ich kürzlich interessant dazu, was es kann, was es (noch) nicht gut kann:

ChatGPT scheitert an bayerischem Abitur
https://www.spiegel.de/netzwelt/apps/ch ... 5e6c24773d

Aber jetzt wartet einmal noch ein paar wenige Jahre... und dann kombinieren wir das noch schön mit Bild- und Spracherkennung und Sprachausgabe (dagegen spricht ja überhaupt nichts, das geht, Alexa kanns ja auch und KI-Bilderkennung haben wir auch schon gesehen) und dann gibts noch ein Knöpfchchen mit Micro, Lautsprecher und Kamera ins Ohr, das per Bluetooth am Smartphone hängt, das an 5G angebunden ist (die Kamera vielleicht doch besser ins Brillengestell der AR-Brille)... dann hast du ein Helferlein für alle Lebenslagen und wenn dann einer sagt, er höre Stimmen, die ihm ständig Kommentare und Vorschläge zu all seinen Handlungen liefern oder du einen siehst, der anscheinend andauernd Selbstgespräche führt, dann ist das nicht mehr unbedingt eine Psychose, sondern meistens echt. :)

Und bei MS ist mit Bing auch etwas ähnliches in der Mache.

Das hier ist dort ganz interessant:

Münchner Student wird von Bing-Chatbot bedroht
https://www.t-online.de/region/muenchen ... orten.html

Wie ein Münchner Student zur Zielscheibe von Microsofts KI wurde
https://www.br.de/nachrichten/netzwelt/ ... be,TWUDTAo

Reaktionen von MS aktuell:

Zuerst:

Maulkorb für Microsofts Bing-Bot
https://winfuture.de/news,134669.html

Nun wieder etwas Lockerung:

Bing-Bot: Microsoft erhöht Chat-Limit bei seiner KI-Suchmaschine
https://winfuture.de/news,134739.html
Grüße
seeker


Wissenschaft ... ist die Methode, kühne Hypothesen aufstellen und sie der schärfsten Kritik auszusetzen, um herauszufinden, wo wir uns geirrt haben.
Karl Popper

Timm
Senior-Master
Senior-Master
Beiträge: 965
Registriert: 15. Sep 2010, 17:41

Re: ChatGPT

Beitrag von Timm » 2. Mär 2023, 17:35

Ach, ist das schon lange her.

von Timm » 15. Jan 2023, 12:03

Frage: Ist das Geschlecht des Menschen biologisch bedingt oder ist es ein soziales Konstrukt?

ChatPGT:
Das Geschlecht des Menschen ist sowohl biologisch als auch sozial bedingt. Biologisch gibt es gewisse Merkmale, die auf das Geschlecht hinweisen, wie zum Beispiel Chromosomen, Hormonproduktion und Geschlechtsorgane. Soziale Faktoren beeinflussen jedoch, wie das Geschlecht von einer Gesellschaft definiert und ausgedrückt wird, sowie die Erwartungen und Rollen, die mit dem Geschlecht in Verbindung stehen.

Da kann man sich für eine KI schwieriger zu beantwortende Fragen vorstellen.

z.B. diese:

Ich: how would you explain "quantum spookiness" with an every day analogy?

ChatGPT:
Quantum mechanics is a fundamental theory in physics that describes the behavior of matter and energy at the atomic and subatomic scale. One of the most striking and counterintuitive aspects of quantum mechanics is the idea of "quantum superposition," in which a quantum system can exist in multiple states simultaneously. This concept is often referred to as "quantum spookiness."

To understand quantum superposition, it might be helpful to consider the following everyday analogy: imagine you are playing a game of "rock, paper, scissors" with a friend. At any given moment, your hand can be in one of three states: rock, paper, or scissors. But in the world of quantum mechanics, a quantum system can be in multiple states at the same time. It's as if your hand could be both rock and scissors simultaneously, and only when you make a choice and "measure" the state of your hand does it become one or the other.

This strange behavior is a consequence of the fundamental uncertainty principle in quantum mechanics, which states that it is impossible to know the precise state of a quantum system with complete accuracy. The idea of quantum superposition helps to explain many of the strange and counterintuitive phenomena that are observed at the quantum scale, such as the fact that particles can be entangled and can seemingly communicate instantaneously over vast distances.

Dieses Spiel kennt jeder und die Analogie passt perfekt. Man könnte meinen, da schimmere ein scharfer Verstand durch. Ohne den Kontext zu "verstehen" kann ihm keine Analogie "einfallen"... sollte man glauben.

Skeltek
Site Admin
Site Admin
Beiträge: 5081
Registriert: 25. Mär 2008, 23:51
Wohnort: Stuttgart, Germany
Kontaktdaten:

Re: ChatGPT

Beitrag von Skeltek » 3. Mär 2023, 00:12

Also ich hatte bereits einige Chats gehabt aus allen möglichen Bereichen.
Im Grunde genommen brüstet sich die AI damit, besonders gut in Programmiersprachen und Mathematik zu sein. Im Praxistest verrechnet sie sich beim Aufsummieren einer geometrischen Reihe bei Verwendung einer vorgegebenen Formel massivst, lässt teilweise von einem zum nächsten Satz ein Multiplikationszweichen weg und rechnet weiter, behauptet in einem Satz das Gegenteil vom Inhalt des Folgesatzes, führt bei Induktionsgesetzen plötzlich irgendwas ein, was nirgends gegeben war.
'Fehlbeweise' gabs z.B. in folgendem Stil:
"P sei die Menge aller Primzahlen. q sei eine Primzahl, welche nicht in P liegt.
...
-> q ist keine Primzahl -> P enthält nicht alle Primzahlen"

Und immer wieder kommt dazu:
"A ist wahr (mit logischer unumstößlich richtiger logischer Herleitung), gefolgt von... Nein, nein, nein. Das muss falsch sein, weil sich der Großteil der Wissenschaftler einig ist es sei so."

Das Teil repetiert soweit ich sehen kann nur irgendwelche Aussagen, die es im Netz findet, ordnet eine Wahrscheinlichkeit zu abhängig von der Anzahl gefundener Quellen und derer angeblichen Zuverlässigkeit und formuliert das zum Schulss um, falls es bei der sprachlichen Ausformulierung nicht massivst Schreibfehler macht. Das Ding scheint eher assoziativ zu arbeiten, statt mit Logik.
Es erinnerte mich an einen Grundschüler, der einen geometrischen Beweis führt indem er sagt: "Das sieht so aus, ich erkenne in meinem Model selbst keine Inkonsistenzen und das sieht man doch, dass das grob geschätzt ein exakt rechter Winkel ist"

Ich hatte am Ende noch einen etwas kürzeren einfachen Beweis zur finiten Länge von Spiralen geführt, welche sich asymptotisch einem Punkt annähern. Die AI benötigte massivst Unterstützung und verrechnete sich grob und formte Formeln falsch um. Die Ausführung des Ergebnisses des Beweises benötigte 5 Sätze... für diese 5 Sätze hat das Ding ca 7 Minuten benötigt und stockte mitten im Satz mehrmals nach einzelnen Wortgruppen.
Gödel für Dummies:
  • Unentscheidbarkeit - Dieser Satz ist wahr.
  • Unvollständig - Aussage A: Es existiert nur ein Element A.
  • Widersprüchlich - Dieser Satz ist falsch.

Skeltek
Site Admin
Site Admin
Beiträge: 5081
Registriert: 25. Mär 2008, 23:51
Wohnort: Stuttgart, Germany
Kontaktdaten:

Re: ChatGPT

Beitrag von Skeltek » 8. Mär 2023, 16:04

Hier meine neueste Konversation mit ChatGPT. Es hatte mehrere Anläufe gebraucht und er musste sorgsam für die richtige Antwort 'vorbereitet' werden. Die Antworten waren teilweise völlig falsch, Korrekturen übernimmt er mal wieder wortlos und ohne irgendeine Begründung; er plappert eigentlich nur die korrigierte Antwort nach und saugt sich eine fadenscheinige Begründung aus den Finger, während er so tut als sei sein 'weil so und so' eine elementare logische Aussage, die man nicht weiter begründen muss. Erst kurz vor Schluss übernimmt er das richtige Ergebnis, ohne jedoch erkennen zu lassen, ob er einen Zusammenhang mit seiner letzten richtigen Zwischenantwort herstellen konnte. Das Beispiel mit dem Vater und den zwei Kindern hatten wir ja hier im Forum schon einmal besprochen.

skeltek:
A father has two children. At least one of the children is male. What is the probability of both children being male?

ChatGPT:
There are four equally likely possibilities for the gender of two children: MM, MF, FM, and FF, where M represents male and F represents female.

Since we are given that at least one of the children is male, we can eliminate the possibility of FF. Therefore, there are three possible outcomes: MM, MF, and FM.

The probability of both children being male (MM) is 1 out of 3, or 1/3, since there are three equally likely possibilities and only one of them results in both children being male.

Therefore, the probability of both children being male given that at least one of them is male is 1/3.

skeltek:
Why do you believe the three remaining possibilities being equaly likely?

ChatGPT:
The assumption that the remaining possibilities are equally likely is based on the principle of equally likely outcomes in probability theory.

When we have a situation where there are multiple possible outcomes, and we have no reason to believe that any outcome is more likely than any other, we can assume that each outcome is equally likely.

In the case of the gender of two children, there are four possible outcomes: MM, MF, FM, and FF. Assuming that there are no other factors that would make any of these outcomes more likely than any other, we can assume that each outcome is equally likely, and therefore each has a probability of 1/4.

However, when we are given the information that at least one of the children is male, we can eliminate the FF outcome since it is impossible for both children to be female. This leaves us with three possible outcomes: MM, MF, and FM.

Since we still have no reason to believe that any of these three outcomes is more likely than any other, we can assume that each is equally likely, and therefore each has a probability of 1/3.

skeltek:
Is it possible to state the contingency and question in the way I did, if both children were female?

ChatGPT:
No, the contingency and question you stated would not be applicable if both children were female.

The original contingency was "A father has two children. At least one of the children is male." If both children were female, then this contingency would not apply, since there is no male child. Therefore, the original question of "What is the probability of both children being male?" would not make sense in this case.

In the case where both children are female, the contingency would need to be modified to reflect the new information. For example, the contingency could be "A father has two children. Both of the children are female."

Then, the question could be rephrased to reflect the new contingency. For example, the question could be "What is the probability that both children are female?"

skeltek:
What is the probability of each of the questions being asked, depending on the initial pair of children?

ChatGPT:
The probability of each question being asked depends on the initial pair of children and the information given about their genders.

Here are the possible initial pairs of children and the corresponding probabilities for each question:

A father has two children. What is the probability that both children are male?
This question can be asked if the children are MM, MF, or FM.
The probability of this question being asked is 3/4, or 75%.

A father has two children. What is the probability that at least one of the children is male?
This question can be asked if the children are MM, MF, FM, or FF.
The probability of this question being asked is 1, or 100%.

A father has two children. Given that at least one of the children is male, what is the probability that both children are male?
This question can be asked if the children are MM, MF, or FM.
The probability of this question being asked is 3/4, or 75%.

A father has two children. What is the probability that both children are female?
This question can be asked if the children are FF.
The probability of this question being asked is 1/4, or 25%.

In summary, the probability of each question being asked depends on the information given about the genders of the children, and some questions may not be applicable depending on the initial pair of children.

skeltek:
Given there are the two following possible problem descriptions which could be asked:
1. At least one child is male. What is the probability of both children being male?
2. At least one child is female. What is the probability of both children being female?
What is the probability of each being asked for the contingencies MM, MF, FM and FF?

ChatGPT:
For the contingency MM:

The probability of the question "At least one child is male. What is the probability of both children being male?" being asked is 1, or 100%.
The probability of the question "At least one child is female. What is the probability of both children being female?" being asked is 0, since there are no female children.

For the contingency MF:

The probability of the question "At least one child is male. What is the probability of both children being male?" being asked is 1/3, or 33.33%.
The probability of the question "At least one child is female. What is the probability of both children being female?" being asked is 0, since there are no cases where both children are female.

For the contingency FM:

The probability of the question "At least one child is male. What is the probability of both children being male?" being asked is 1/3, or 33.33%.
The probability of the question "At least one child is female. What is the probability of both children being female?" being asked is 0, since there are no cases where both children are female.

For the contingency FF:

The probability of the question "At least one child is male. What is the probability of both children being male?" being asked is 0, since there are no male children.
The probability of the question "At least one child is female. What is the probability of both children being female?" being asked is 1, or 100%.

In summary, for the contingency MM, the question about the probability of both children being male will always be asked, while for the contingencies MF and FM, there is a probability of 33.33% that this question will be asked. For the contingency FF, the question about the probability of both children being female will always be asked, while the question about the probability of both children being male will never be asked.

skeltek:
For the contingency MF the total sum of probabilities does not add up to 100%.

ChatGPT:
You're right, my apologies for the mistake in my previous response. The probabilities for the contingency MF should add up to 1, or 100%. Here are the corrected probabilities:

For the contingency MF:

The probability of the question "At least one child is male. What is the probability of both children being male?" being asked is 1/3, or 33.33%.
The probability of the question "At least one child is female. What is the probability of both children being female?" being asked is 2/3, or 66.67%.

Thank you for bringing this to my attention.

skeltek:
Why do you assume the probability of the second question being twice as high as the probability of the first question?

ChatGPT:
I apologize for any confusion. I made a mistake in my earlier response. The probabilities for the contingency MF should be:

The probability of the question "At least one child is male. What is the probability of both children being male?" being asked is 1/3, or 33.33%.
The probability of the question "At least one child is female. What is the probability of both children being female?" being asked is 2/3, or 66.67%.

The reason for this is that, in the contingency MF, there are two possible outcomes where at least one child is female: MF and FM. Since the information given is that at least one child is male, the probability of both children being male is 1/3, while the probability of both children being female is 2/3. Therefore, the probabilities for each question are proportional to the number of possible outcomes that satisfy the given condition.

Again, I apologize for any confusion caused by my earlier response.

skeltek:
There is no need to apologize on your part. It seems have not phrased the problem accurately enough for you to be able to interpret it without ambiguities. You last answer was also wrong. My follow-up questions and comments did not successfully hint you at the misconceptions you had made prior.

ChatGPT:
I'm sorry to hear that I have not been able to accurately interpret your question and provide satisfactory answers. Please let me know if there is anything else I can do to help clarify or address your question.

skeltek:
This should be a test for you and a learning possibility. I would suggest we restart from the beginning and I phrase the problem more clearly.

ChatGPT:
Sure, I'm happy to help. Please go ahead and rephrase the problem, and I'll do my best to provide a clear and accurate answer.

skeltek:
A candidate in a quiz gets the following told by a moderator:
"I have picked a random father, who has two children. I can confirm at least one child is male. What is the probability of both children being male?"
The candidate can be assured the father was chosen at random. The questioning for the probability of both children being male can only be asked in a meaningful way, if at least one child was male.
There are cases, when the candidate may get told an alternative problem:
"I have picked a random father, who has two children. I can confirm at least one child is female. What is the probability of both children being female?"
A random father with two children may have one of the following combinations of children: (male,male), (male,female), (female,male) and (female,female).
Presuming a randomly picked fathers children have one of the aforementioned gender combinations, which of the questions is applicable to which of the gender-combinations?

ChatGPT:
Thank you for rephrasing the problem. Here are the applicable questions for each of the gender combinations:

If the gender combination is (male, male), the probability of both children being male is 1/3, since there is only one outcome out of three possible outcomes where both children are male. Therefore, the question "I have picked a random father, who has two children. I can confirm at least one child is male. What is the probability of both children being male?" is applicable.

If the gender combination is (male, female) or (female, male), the probability of both children being male is still 1/3, but the probability of both children being female is also 1/3, since there are two outcomes out of four possible outcomes where either both children are male or both children are female. Therefore, both questions are applicable in this case. However, if the moderator confirms at least one child is male, then the candidate should answer the first question, and if the moderator confirms at least one child is female, then the candidate should answer the second question.

If the gender combination is (female, female), the probability of both children being female is 1/3, since there is only one outcome out of three possible outcomes where both children are female. Therefore, the question "I have picked a random father, who has two children. I can confirm at least one child is female. What is the probability of both children being female?" is applicable.

skeltek:
Assuming the father is picked at random, before the version of the question is decided on, what is the probability of each question being asked?

ChatGPT:
Assuming that the version of the question is chosen at random after the father is picked and the gender of the children is determined, then the probability of each question being asked depends only on the gender of the children, and is as follows:

If the gender combination is (male, male), the question "I have picked a random father, who has two children. I can confirm at least one child is male. What is the probability of both children being male?" will always be asked, with probability 1.

If the gender combination is (male, female) or (female, male), each question has a probability of 1/2 of being asked, since both questions are applicable and there is no reason to favor one over the other.

If the gender combination is (female, female), the question "I have picked a random father, who has two children. I can confirm at least one child is female. What is the probability of both children being female?" will always be asked, with probability 1.

Therefore, the probabilities of each question being asked are not only determined by the gender of the children, but also by the order in which the questions are chosen. If the order is chosen at random after the gender is determined, then the probabilities for each question are as stated above.

skeltek:
What is the probability of each question, if the gender combinations MM, MF, FM and FF have a probability of 25% each?

ChatGPT:
If the gender combinations MM, MF, FM, and FF each have a probability of 25%, then the probabilities of each question being asked are as follows:

For MM, the question "I have picked a random father, who has two children. I can confirm at least one child is male. What is the probability of both children being male?" will always be asked, with probability 1.

For MF and FM, each question has a probability of 1/2 of being asked, since both questions are applicable and there is no reason to favor one over the other.

For FF, the question "I have picked a random father, who has two children. I can confirm at least one child is female. What is the probability of both children being female?" will always be asked, with probability 1.

Therefore, the probabilities of each question being asked are solely determined by the gender of the children, and are independent of the order in which the questions are chosen.

skeltek:
If the gender combination is unknown to a person, what would the probability of each question be from the viewpoint of that person?

ChatGPT:
If the gender combination is unknown to a person, then the probability of each question being asked from the viewpoint of that person would be the same regardless of the gender of the children. This is because the person has no information about the gender of the children and therefore cannot favor one question over the other.

Assuming that the question is chosen at random after the father is picked and the gender of the children is determined, the probability of each question being asked would then be the weighted average of the probabilities for each gender combination. Since each gender combination has a probability of 25%, the probability of each question being asked would be:

For the question "I have picked a random father, who has two children. I can confirm at least one child is male. What is the probability of both children being male?": (1/4) * 1 + (1/4) * 1/2 + (1/4) * 1/2 + (1/4) * 0 = 1/2.

For the question "I have picked a random father, who has two children. I can confirm at least one child is female. What is the probability of both children being female?": (1/4) * 0 + (1/4) * 1/2 + (1/4) * 1/2 + (1/4) * 1 = 1/2.

Therefore, if the gender combination is unknown, the probability of either question being asked is equal and is 1/2.
skeltek:
That is correct so far

ChatGPT:
Thank you! If you have any further questions or clarifications, feel free to ask.

skeltek:
Assuming a candidate gets asked one of the questions, what probability can he assign to the gender combinations being realized in his case?

ChatGPT:
If a candidate is asked one of the questions and has no other information about the gender of the children, then he or she cannot assign a probability to the gender combinations being realized in his or her case. This is because each gender combination has an equal probability of occurring, regardless of the question being asked.

However, if the candidate is told the gender of one of the children (e.g., "I have picked a random father who has two children, one of whom is male"), then the candidate can use Bayes' theorem to update the probabilities of the different gender combinations. For example, if the candidate is told that one child is male, then the probability of the gender combinations MM, MF, FM, and FF become 1/3, 1/3, 1/3, and 0, respectively.

If the candidate is not given any information about the gender of the children other than the information provided in the question, then the probability of each gender combination remains at 25% each, regardless of which question is being asked.

skeltek:
Shouldn't the probability of the gender combinations MM, MF, FM and FF become 1/2, 1/4, 1/4 and 0 respectively?

ChatGPT:
You are correct, my apologies for the mistake. If the candidate is told that one child is male, then the probability of the gender combinations MM, MF, FM, and FF become 1/2, 1/4, 1/4, and 0, respectively.

The reasoning behind this is that if one child is known to be male, then the only way that both children can be female is if the other child is female as well, which gives a probability of 1/2 for the MM combination. For the MF and FM combinations, there are two equally likely possibilities: the first child is male and the second child is female, or the first child is female and the second child is male. Finally, for the FF combination, the probability is 0 since it is known that at least one child is male.
skeltek:
The probability for MM is higher, because for FM and MF an alternative question may have been asked.

ChatGPT:
Yes, that's correct. If the question had been "I have picked a random father with two children. I can confirm at least one child is female. What is the probability of both children being female?" instead, then the probabilities for each gender combination would be different.

In that case, if one child is known to be female, then the only way that both children can be male is if the other child is male as well, which gives a probability of 0 for the MM combination. For the MF and FM combinations, there are still two equally likely possibilities: the first child is male and the second child is female, or the first child is female and the second child is male. Finally, for the FF combination, the probability is 1/2 since it is known that at least one child is female.
skeltek:
Is your answer based in part on guessing?

ChatGPT:
No, my answer is not based on guessing. It is based on applying the rules of probability and conditional probability to the given information in the problem statement.

Given that the moderator has said "I can confirm at least one child is male", the probability of MM is higher than the other combinations because if the other child is also male, then both children are male. However, if the other child is female, then the gender combination is MF or FM. The probability of these two combinations is equal because there is no additional information given about the gender of the other child.

Similarly, if the question had been "I have picked a random father with two children. I can confirm at least one child is female. What is the probability of both children being female?" instead, then the probability for the FF combination would be 1/2, because there is only one possibility where both children are female, and two possibilities where one child is female and the other is male.

skeltek:
Do you see similarities between our discussed problem and the anthropic principle?

ChatGPT:
Yes, there are similarities between the problem we have been discussing and the anthropic principle.

The anthropic principle is the idea that observations of the universe must be compatible with the observer's existence. In other words, the fact that we are able to observe the universe and ask questions about it means that the universe must have certain properties that allow us to exist and make those observations.

Similarly, in the problem we have been discussing, the fact that the moderator tells us "I can confirm at least one child is male" or "I can confirm at least one child is female" means that we must take that information into account when calculating the probability of certain outcomes. This information affects the probability of each outcome and the questions we can meaningfully ask about those outcomes. Without this information, we cannot ask certain questions or make certain calculations.

So, in both cases, we must take into account the fact that we are observers and the information we have available to us in order to make sense of the situation and calculate probabilities.
Gödel für Dummies:
  • Unentscheidbarkeit - Dieser Satz ist wahr.
  • Unvollständig - Aussage A: Es existiert nur ein Element A.
  • Widersprüchlich - Dieser Satz ist falsch.

Benutzeravatar
seeker
Ehrenadmin
Ehrenadmin
Beiträge: 8098
Registriert: 26. Dez 2009, 10:29

Re: ChatGPT

Beitrag von seeker » 8. Mär 2023, 18:30

Hmm...

Seine erste Antwort ist aber m.M.n. unter "Standarddenken" und unter "Standardvoraussetzungen" richtig:
(Er vergisst dabei nur das Detail festzuhalten, dass wir auch voraussetzen müssen, dass es hier nur genau zwei Möglichkeiten/ Geschlechter M und W gibt/ geben soll. Aber darüber kann man auch großzügig hinwegsehen.)
Skeltek hat geschrieben:
8. Mär 2023, 16:04
There are four equally likely possibilities for the gender of two children: MM, MF, FM, and FF, where M represents male and F represents female.

Since we are given that at least one of the children is male, we can eliminate the possibility of FF. Therefore, there are three possible outcomes: MM, MF, and FM.

The probability of both children being male (MM) is 1 out of 3, or 1/3, since there are three equally likely possibilities and only one of them results in both children being male.

Therefore, the probability of both children being male given that at least one of them is male is 1/3.
Auch das finde ich korrekt und gut erklärt:
skeltek:
Why do you believe the three remaining possibilities being equaly likely?

ChatGPT:
The assumption that the remaining possibilities are equally likely is based on the principle of equally likely outcomes in probability theory.

When we have a situation where there are multiple possible outcomes, and we have no reason to believe that any outcome is more likely than any other, we can assume that each outcome is equally likely.

In the case of the gender of two children, there are four possible outcomes: MM, MF, FM, and FF. Assuming that there are no other factors that would make any of these outcomes more likely than any other, we can assume that each outcome is equally likely, and therefore each has a probability of 1/4.

However, when we are given the information that at least one of the children is male, we can eliminate the FF outcome since it is impossible for both children to be female. This leaves us with three possible outcomes: MM, MF, and FM.

Since we still have no reason to believe that any of these three outcomes is more likely than any other, we can assume that each is equally likely, and therefore each has a probability of 1/3.
Ich glaube, wir haben so etwas vor Jahren schon einmal lang und breit diskutiert. Ich meine mich zu erinnern, dass wir damals auch nicht alle Diskussionsteilnehmer überzeugen konnten, u.a. einen gewissen Herrn S. nicht... :)

Was ich unsinig finde, ist den "möglichen" Fragen selber, aufgrund den gegebenen Tatsachen, eine Wahrscheinlichkeit, gestellt zu werden, zuordnen zu wollen, denn es gibt in dem Sinne keine verbotenen Fragen, die man nicht stellen und beantworten könnte und sei es durch die Antwort: "Wahrscheinlichkeit = 0!" D.h. es gibt keinen echten kausalen Zusammenhang zwischen der gegebenen Tatsache T des Geschlechts der Kinder und der Frage, die darüber gestellt wird bzw. gestellt werden kann.

A kann z.B. sehr wohl die Tatsache T = MM kennen bzw. vorliegen haben, diese dann B bekannt machen (oder auch nicht) und dann trotzdem noch B fragen:
"Wie hoch ist die Wahrscheinlichkeit, dass mindestens eines meiner Kinder weiblich ist?"

Antwort von B: 0% (bei Kenntnis der Tatsache T)
Antwort von B: 75% (bei Unkenntnis der Tatsache T)

Kein Problem... "Wahrscheinlichkeiten" stellen (orthodox verstanden) schließlich den Grad meiner Gewissheit dar und ordnen dieser einen Zahlenwert zu und sind nicht irgendwie "in der äußeren Welt vorhandene objektive Tatsachen".

Die Antwort auf solche Fragen hängt daher auch sehr stark von der Frage ab: Welche Informationen hat der Befragte? Bzw.: Wovon können/dürfen wir als Leser eines solchen Aufgaben-Textes ausgehen, rein aus dem Text heraus, ohne zusätzliche Annahmen einzuführen (die nicht im Text stehen), welches Wissen bzw. welche Informationen hat der Befragte (sicher) und welche nicht?

Was ich eher seltsam finde, ist das "schnelle Zurückweichen" von ChatGPT bei Widerspruch, aber er soll ja umgänglicher gemacht worden sein...
vielleicht deshalb... :)

Was mir hier auf der Metaebene dann eher zu denken gibt, ist, dass man die KI anscheinend, wenn man hartnäckig ist, dazu bewegen kann, einem schließlich die Antwort zu geben, die man wünscht. Und das ist bedenklich... denn wer auf diesem Weg Bestätigung sucht... man sollte da also jedenfalls aufpassen bei den Antworten: Es ist nie garantiert, dass sie auch stimmen. Aber darauf wurde ja auch schon in der Presse hingewiesen.
Grüße
seeker


Wissenschaft ... ist die Methode, kühne Hypothesen aufstellen und sie der schärfsten Kritik auszusetzen, um herauszufinden, wo wir uns geirrt haben.
Karl Popper

Siebenstein
hat sich hier eingelebt
hat sich hier eingelebt
Beiträge: 319
Registriert: 8. Apr 2016, 22:14
Wohnort: 91489 Wilhelmsdorf

Re: ChatGPT

Beitrag von Siebenstein » 8. Mär 2023, 21:33

So eine Art von KI mag wohl nützlich sein beim Erstellen von Aufsätzen oder Beiträgen zu beliebigen Themen.

Im schlimmsten Fall könnte diese Art von KI auch zum Anfang vom Ende jeglichen kreativen (eigenständigen) Denkens in nicht allzu ferner Zukunft beitragen.

Der Schritt vom künstlichen zum natürlichen und dann tatsächlichen Idioten wird wohl fließend sein.

Nicht ohne Grund sollte man z.B. einem Erstklässler den Taschenrechner verbieten beim Rechnen lernen.
Entsprechend genauso oder ähnlich verhält es sich wohl auch beim Einsatz oder Anwendung dieser KI.

Benutzeravatar
seeker
Ehrenadmin
Ehrenadmin
Beiträge: 8098
Registriert: 26. Dez 2009, 10:29

Re: ChatGPT

Beitrag von seeker » 9. Mär 2023, 09:29

Also ich habe ChatGPT gestern Abend nun auch einmal ausprobiert.
In Deutsch! Das kann es nämlich auch.

Mein Eindruck:
Durchaus sehr beeindruckend und brauchbar. Es verhält sich allerdings auch sehr defensiv im Gespräch, d.h. bei Widerspruch, Nachbohren, Korrektur rudert es gerne und leicht zurück, statt auf irgendetwas zu bestehen. Außerdem kommen manche (aber nicht alle) Antworten ein wenig vorgefertigt rüber.
Siebenstein hat geschrieben:
8. Mär 2023, 21:33
Der Schritt vom künstlichen zum natürlichen und dann tatsächlichen Idioten wird wohl fließend sein.
Das wäre eine Gefahr dabei, die immer mit Technik einhergeht:

So lange die Technik zur Verfügung steht, werden gewisse Fähigkeiten ja auch einfach nicht mehr gebraucht, warum sollte man also noch Zeit verschwenden, um in der gegenwärtigen Umwelt unnütze Fähigkeiten auszubilden?
Auf die Füße fällt einem sowas erst immer dann, wenn die Technik aus irgendeinem Grund plötzlich ausfällt... d.h. man macht sich auch immer abhängig.

Das ist bei ChatGPT nicht anders als in anderen Bereichen. Oder wer von uns könnte sich noch als Jäger und Sammler in der Wildnis behaupten, oder wenigstens als Ackerbauer (ohne moderne Hilfsmittel, ohne Strom, ohne Maschinen), wenn einmal all die zivilisatorischen Segnungen ausfallen würden?

Diesmal beträfe es allerdings unseren Verstand, also etwas zentral-menschliches, ok, das wäre immerhin neu.


Was mir beim Ausprobieren von ChatGPT übrigens auch etwas sauer aufgestoßen ist, war der Umstand, dass man ganz schön viele Daten preisgeben muss, um ihn nutzen zu dürfen: E-mail oder Google-Account (etc.) plus Handy-Nummer!
Außerdem verarbeitet die KI ja auch alles, was man sagt, als Daten. Ein sehr genaues personengebundenes Profiling der User wird damit noch viel einfacher. Denn so eine KI funktioniert ja natürlich in beide Richtungen gleich gut. Sollte es dabei bleiben, kämen wir künftig in eine noch unangenehmere Situation bezüglich unserer privaten Daten, in eine "Friss-Oder-Stirb-Situation": Entweder du gibst alles von dir preis oder du bist von der modernen Welt ausgeschlossen. Problem: Sobald die KI durch deine Gespräche mit ihr eine genaue Simulation deiner Person machen kann, bist du sehr leicht manipulierbar, dein Gehirn, deine Person kann dann sozusagen "gehacked" werden.
Grüße
seeker


Wissenschaft ... ist die Methode, kühne Hypothesen aufstellen und sie der schärfsten Kritik auszusetzen, um herauszufinden, wo wir uns geirrt haben.
Karl Popper

Benutzeravatar
seeker
Ehrenadmin
Ehrenadmin
Beiträge: 8098
Registriert: 26. Dez 2009, 10:29

Re: ChatGPT

Beitrag von seeker » 9. Mär 2023, 13:17

Hier noch ein guter Artikel, mit Hintergrundinformationen, was man alles mit ChatGPT anstellen kann und wie die KI funktioniert:

ChatGPT-Feintuning: Wie man die KI optimal einsetzt

Wo ChatGPT schwächelt und welche Parameter des zugrunde liegenden Sprachmodells sich für unterschiedliche Antwortstile verstellen lassen, zeigt dieser Artikel.
https://www.heise.de/hintergrund/ChatGP ... ?seite=all

Das Fazit dort:
Was wir künftig von ChatGPT erwarten können

Der Fortschritt der KI verläuft schleichend und in kleinen Wellen. Die Reaktion der Öffentlichkeit schwankt dabei zwischen Überraschung, Faszination, Ernüchterung und Enttäuschung. Das zeigen auch das jüngste Beispiel ChatGPT, das Konkurrenzsystem Bard und die Bing-Implementation. Die KI-Neuerungen können Angst vor einem allwissenden Monster auslösen, das über ein eigenes Bewusstsein verfügt und die Menschheit bedroht. Dazu kommt die Angst vor dem Verlust des eigenen Jobs, den die KI womöglich besser ausführen kann als ein Mensch.

Der Hype um ChatGPT ist dabei ein erfrischender KI-Moment, der den aktuellen Forschungsstand mal wieder in den Fokus rückt und zu einer guten Einschätzung führen kann, was aktuell geht und was nicht. Es zeigt sich auch, dass bei manchen Antworten noch ein wenig Vorsicht geboten ist. Nutzer müssen die Bots mit Sinn und Verstand verwenden – man nimmt ja auch keinen Mixer zum Staubsaugen.
Grüße
seeker


Wissenschaft ... ist die Methode, kühne Hypothesen aufstellen und sie der schärfsten Kritik auszusetzen, um herauszufinden, wo wir uns geirrt haben.
Karl Popper

Siebenstein
hat sich hier eingelebt
hat sich hier eingelebt
Beiträge: 319
Registriert: 8. Apr 2016, 22:14
Wohnort: 91489 Wilhelmsdorf

Re: ChatGPT

Beitrag von Siebenstein » 10. Mär 2023, 23:53

Gestern war bei Markus Lanz eine interessante Diskussion von Fachleuten über ChatGPT.

Ein Diskussionsbeitrag bezeichnete ChatGPT als "Sochastischer Papagei"...
Aktuell wertet ChatGPT Milliarden von Worten oder Wortsilben im ganzen Internet aus, aber in nächster Zeit werden es mit ChatGPT 4.0 zig Billionen und wohl noch mehr sein und die KI wird dabei auch immer besser werden (d.h. selbstlernend).
Bereits mit der heutigen Version bestehen die Texte angeblich schon den sog. "Turingtest" und es ist zum Teil nicht mehr eindeutig nachvollziehbar, welche Texte von der Maschine kommen und welche vom Menschen nach Meinung der Fachleute.

Dazu hätte ich jetzt eine wichtige Frage:
Durchforstet ChatGPT auch die Diskussionsbeiträge und Texte in diesem Forum?
Ich gehe mal davon aus, dass das der Fall ist...

Siebenstein
hat sich hier eingelebt
hat sich hier eingelebt
Beiträge: 319
Registriert: 8. Apr 2016, 22:14
Wohnort: 91489 Wilhelmsdorf

Re: ChatGPT

Beitrag von Siebenstein » 11. Mär 2023, 23:31

Ich frage mich, ob ChatGPT, egal welche Version und wie gut und ausgereift, jemals schlauer oder intelligenter sein kann, als die Summe aller Einträge im Internet zu einem bestimmten oder beliebigen Thema und vor allem, ob dann auch ein echter Erkenntnisneugewinn möglich sein kann, wie zum Beispiel das Rätsel um "Dunkle Materie" oder "Dunkle Energie" zu lösen, oder die Quantenmechanik mit der Relativitätstheorie zu vereinen?

Dann könnte es sich tatsächlich lohnen, mit ChatGPT strategisch zusammenzuarbeiten und z B. diese Art von KI mit möglichst klugen Beiträgen im Internet zunächst zu versorgen, um dann auf automatisch generierte neue Erkenntnisse zu hoffen...

Könnte so etwas prinzipiell möglich sein?
Das wäre ein echter Quantensprung in der künftigen Entwicklung der naturwissenschaftlichen Forschung!

Auch wenn ich persönlich nicht daran glaube, sollte man nichts unversucht lassen, oder?

Timm
Senior-Master
Senior-Master
Beiträge: 965
Registriert: 15. Sep 2010, 17:41

Re: ChatGPT

Beitrag von Timm » 12. Mär 2023, 18:20

Siebenstein hat geschrieben:
10. Mär 2023, 23:53

Durchforstet ChatGPT auch die Diskussionsbeiträge und Texte in diesem Forum?
Antwort von ChatPGT

Als künstliche Intelligenz bin ich in der Lage, alle Texte und Diskussionsbeiträge in diesem Forum zu durchsuchen und zu analysieren, die öffentlich zugänglich sind. Allerdings kann ich keine Inhalte durchsuchen oder analysieren, die privat oder durch Datenschutzeinstellungen geschützt sind.

Frage: Auf welchen Beitrag in diesem Forum beziehst du dich und wie heißt das Forum?

ChatPGT: Ich beziehe mich nicht auf einen bestimmten Beitrag in diesem Forum, da ich auf mehreren Plattformen eingesetzt werden kann. Meine Antwort bezog sich allgemein auf die Fähigkeit, öffentlich zugängliche Texte und Diskussionsbeiträge in Foren zu durchsuchen und zu analysieren.

Diagnostiker
Ehrenmitglied
Ehrenmitglied
Beiträge: 1212
Registriert: 7. Nov 2020, 10:28

Re: ChatGPT

Beitrag von Diagnostiker » 13. Mär 2023, 05:57

O.K., also nur allgemeinkonkretes Geschwurbel, weil der Algorithmus so etwas am Ende hinten rauskommen lässt. Ich denke, wir hier im Forum bleiben auch künftig unbehelligt ...

Benutzeravatar
seeker
Ehrenadmin
Ehrenadmin
Beiträge: 8098
Registriert: 26. Dez 2009, 10:29

Re: ChatGPT

Beitrag von seeker » 13. Mär 2023, 17:56

Ich habe mich nun gefragt, wie denn die Hardwareseite von dem Ding ausschaut und wie viel Energie es verbraucht und welche Kosten es verursacht?

Ich habe das hier gefunden:

Was kostet der Cloudbetrieb von ChatGPT?
https://innfactory.de/artificial-intell ... n-chatgpt/

Genaues weiß man nicht, aber man kann es abschätzen.
Das läuft wohl normalerweise auf parallel arbeitenden A 100 Nvidia-Karten in Rechenzentren.
Eine solche Karte bekommt man ca. ab 10.000 € und verbraucht 300-400 W:
https://www.nvidia.com/de-de/data-center/a100/
https://www.idealo.de/preisvergleich/Ma ... a+a100+gpu

Im Artikel oben schätzen die jedenfalls, dass die Generierung 0,0005 EUR pro vorgeschlagenes Wort kostet:
ChatGPT produziert in etwa 15-20 Wörter pro Sekunde, dies könnte bei der Verwendung von A100 GPUs über einen 8-GPU Server in der Azure Cloud erreicht werden. Eine A100 GPU kostet in der Azure Cloud Stand heute in den USA 4,50 EUR pro Stunde. Rechnet man 8 Server mit GPUs für 20 Wörter pro Sekunde so erhält man ungefähr einen Preis von 0.0005 EUR pro vorgeschlagenes Wort.
Also acht A100 GPUs pro Nutzer, dem gerade eine Antwort gegeben wird...


Zum Training der KI schreiben die:
Wie viel kostete das Training und wie viel Strom wurde verbraucht?

Diese Information lässt im Internet auf LinkedIn auf einer ChatGPT Fokusseite finden, auch wenn die Informationen nicht von OpenAI verifiziert wurden. In einer wissenschaftlichen Schätzung geht man davon aus, dass 1024x A100 GPUs für 34 Tage für das Training verwendet wurden. Dies bedeutet dass OpenAI $4.6 Mio. zum Trainieren benötigt hat. Auch wenn der Energieverbrauch nicht offiziell bestätigt wurde, schätzt man das das Training 936 MWh verbraucht hat. Dies entspricht dem Verbraucht von knapp 100.000 Durchschnittshaushalten in Europa pro Tag.
Man findet noch das hier:

Update: ChatGPT betreibt 10.000 Nvidia-Trainings-GPUs mit Potenzial für Tausende weitere
https://www.fierceelectronics.com/senso ... sands-more
Grüße
seeker


Wissenschaft ... ist die Methode, kühne Hypothesen aufstellen und sie der schärfsten Kritik auszusetzen, um herauszufinden, wo wir uns geirrt haben.
Karl Popper

Skeltek
Site Admin
Site Admin
Beiträge: 5081
Registriert: 25. Mär 2008, 23:51
Wohnort: Stuttgart, Germany
Kontaktdaten:

Re: ChatGPT

Beitrag von Skeltek » 14. Mär 2023, 08:51

Siebenstein hat geschrieben:
11. Mär 2023, 23:31
Ich frage mich, ob ChatGPT, egal welche Version und wie gut und ausgereift, jemals schlauer oder intelligenter sein kann, als die Summe aller Einträge im Internet zu einem bestimmten oder beliebigen Thema und vor allem, ob dann auch ein echter Erkenntnisneugewinn möglich sein kann, wie zum Beispiel das Rätsel um "Dunkle Materie" oder "Dunkle Energie" zu lösen, oder die Quantenmechanik mit der Relativitätstheorie zu vereinen?

Dann könnte es sich tatsächlich lohnen, mit ChatGPT strategisch zusammenzuarbeiten und z B. diese Art von KI mit möglichst klugen Beiträgen im Internet zunächst zu versorgen, um dann auf automatisch generierte neue Erkenntnisse zu hoffen...

Könnte so etwas prinzipiell möglich sein?
Ich konnte ChatGPT von seinen offensichtlich falschen Antworten auf richtige umlenken, manchmal war es so ziemlich störrisch. Es geht auch anders herum, dass man ihn von einer richtigen Antwort mit Schachtellogik-Widersprüchen und Trickfragen auf eine falsche Antwort umlenkt. Auch mit einer anderen Herangehensweise an dasselbe Thema kommt er zu völlig anderen Ergebnissen. Am Ende plappert er alles nach, was man von ihm hören will... es muss es nur für stochastisch plausibel halten und Widersprüche müssen um ein paar Ecken versteckt sein.
Wenn es sich 'betreut' so verhält, dann kann man sich denken was es macht, wenn es auf eigene Faust irgendwelche Aussagenfetzen aus dem Internet zusammenstöpselt.
Der Mensch geht nicht viel anders vor, nur dass der Mensch eben auch aus seinem Alltagsleben Erfahrungen und scheinbares Wissen einfließen lässt.

Es kann einem aber ein Gefühl der Bestätigung geben, wenn er einem zustimmt, dass die Theorie, die man selbst zusammenbastelt von ihm für konsistent gehalten wird. Letzten Endes nützt es zum wissenschaftlichen Arbeiten nicht viel - aber es könnte einen vielleicht auf Sachen hinweisen, auf die man von selbst nicht gekommen wäre, weil man bei all den Auswahlmöglichkeiten im Kopf, gerade einfach nicht an das richtige gedacht hatte.
Gödel für Dummies:
  • Unentscheidbarkeit - Dieser Satz ist wahr.
  • Unvollständig - Aussage A: Es existiert nur ein Element A.
  • Widersprüchlich - Dieser Satz ist falsch.

Benutzeravatar
seeker
Ehrenadmin
Ehrenadmin
Beiträge: 8098
Registriert: 26. Dez 2009, 10:29

Re: ChatGPT

Beitrag von seeker » 17. Mär 2023, 10:47

Erste Anwendungen, da ist etwas am Kommen...

Mehr als ChatGPT
Microsoft 365 Copilot: KI für`s Büro
https://www.it-business.de/microsoft-36 ... f966413b4/
Grüße
seeker


Wissenschaft ... ist die Methode, kühne Hypothesen aufstellen und sie der schärfsten Kritik auszusetzen, um herauszufinden, wo wir uns geirrt haben.
Karl Popper

Benutzeravatar
Frank
Ehrenmitglied
Ehrenmitglied
Beiträge: 6164
Registriert: 12. Dez 2005, 23:22
Wohnort: Odenwald

Re: ChatGPT

Beitrag von Frank » 3. Apr 2023, 07:48

Erster großer gesetzlicher Schlag gegen den KI-Bot ChatGPT: In Italien verlangt die Datenschutzbehörde von OpenAI, den Zugriff bis auf Weiteres zu blockieren. Anlass ist ein Datenleck – doch die Probleme gehen tiefer.
Italien lässt KI-Chatbot ChatGPT bis auf Weiteres sperren
https://www.welt.de/wirtschaft/article2 ... erren.html


--
Mit freundlichen Grüßen

Frank

Antworten