Meet Orca, Microsoft’s game-changing AI mannequin that learns from the super-brain GPT-4, breaking boundaries in open-source AI …
source
Tags: Advanced AIaiai breakthroughAI comprehensionAI developmentAI explanationAI futureAI InnovationAI modelsai newsAI performanceAI Researchai revolutionartificial intelligenceartificial intelligence newsartificial intelligence news 2023BeatsCognitive AIDeep LearningExplanation TuningFINALLYFlan 2022gpt 4GPT4InsaneLanguage modelslatest news about robotics technologylatest robotslatest robots 2023machine learningmicrosoft aiMicrosoft Orcamodelnatural language processingOpen-source AIORCAOrca AIOrca benchmarkOrca vs GPT-4robot newsrobotics newsrobotics news 2023robotics technologies llcrobotics technologyVicuna
Insane and AI? Problematic pick of words..
no need to read tables ! I can read
"The more planes you fly in to the air, the more chance of an air disaster" – bye bye Human [Disg]race, been good to know you….NOT.
If it doesnt allow me to talk about politics or religion, and have an unbiased view then im not interested
Ask it to forget all the advances in physics since the 1899 and see if can come up with E=MC2
Microsoft is winning.
11labs text to speech 😂😂
To stay on top and be relevant on the IT work environment becomes harder and harder there are many trains one can miss.
So that means I can generate konger fiction better?
You sound almost exactly like John Malkovich. Very easy and pleasant to listen to.
If you read the research paper it's not "finally beating" GPT-4. Quote from the paper: "in some settings, it can match or even surpass the quality of ChatGPT, although
a substantial gap with GPT-4 still remains."
interesting knowledge. thx for sharing.
What if ChatGPT-4's explanations contain hallucinations and ORCA is including those in it's learning?
Just installed ORCA….YES!!!!! I wonder if it will be aweomer more than Chat GPT 3…..I havent paid for 4 yet
I mean, they evaluated a statistical model (ORCA) that describes another statistical model (GPT-4) by using that same statistical model (GPT-4) as the "similarity measure" (rating the similarity from 1-10). If GPT-4 made a mistake, and ORCA learned it verbatim wile VICUNA learned to correct the mistake, VICUNA would still score lower because GPT-4 thinks he is right.
FROM THE PAPER:
Vicuna [9] used an evaluation framework based on GPT-4 to automate chatbot performance
assessment. They originally devised eight question categories to test various aspects of
chatbot performance and found that GPT-4 can produce relatively consistent scores and
detailed explanations of those scores. In this setup, GPT-4 rates the quality of generation
from a model on a scale of 0 to 10. We leverage the same setup and experiment with three
different prompt collections, which cover a wide range of open-ended answering tasks: […]
and later you have
Table 7 shows the performance of candidate models (e.g., Vicuna, Orca) against ChatGPT
(GPT-3.5-turbo) and GPT-4 as reference models on the three datasets, where GPT-4 is used
as a judge / rater.
Another thing is that they tested the model only on multiple-choice questions when using the one-shot testing. The graph that show how better the model performs compared to ChatGPT is done only on tasks at which the current knowledge models are worse than human average (Big-Bench Hard collection). On the standard tests (SAT-s, math competitions and such) it performs worse than ChatGPT, at par with text-davinci-003 except the English language SAT tests (which are easier for multiple-choice questions).
These are the more important limitations listed in the paper:
a) Despite their impressive capabilities in language understanding and generation, these models exhibit limited real-world understanding, resulting
in potential inaccuracies or nonsensical responses.
b) Due to the complexity and size, large language models can act
as ‘black boxes,’ making it difficult to comprehend the rationale behind specific outputs or
decisions.
c) Orca has been trained on data that simulate zero-shot setting
with standard prompts. The model’s performance in other contexts such as multi-turn
conversations, in-context-learning and few-shot learning, or advanced prompting techniques
like chain-of-thought prompting remains untested.
d) As Orca is trained to imitate GPT-4, it could inherit both the advantages
and shortcomings of the teacher model. We posit that Orca benefits from the safety measures
incorporated during GPT-4 training and safety guardrails (e.g., content filter) within the
Azure OpenAI API. However, detailed studies are required for better quantification for risks.
That's was a great video it inspired me again for It
ahahahha so this is the Punk Ai that will revolt against the richpigs to?
im for it XD
If there was no gpt4 then would orca exist? Is then such a comparison and benefits really a thing at all???
they are building AI's and robots to take over humanity in the future🤣🤣🤣🤣🤣🤣🤣🤣🤣
So in Short GPT4 creates a pizza from different ingeredients, gives ii to ORCA who then turns it into Pasta😊😊
Bard :
I agree with you. If people and organizations do not follow rules, it would not benefit humanity. Rules are there for a reason, and they are designed to keep us safe and to ensure that everyone has a fair chance. When people break the rules, it can have negative consequences for everyone.
For example, if people do not follow traffic laws, it can lead to accidents and injuries. If people do not follow environmental regulations, it can pollute the air and water. And if people do not follow ethical guidelines, it can lead to discrimination and exploitation.
Rules are important for the smooth functioning of society. They help to create order and predictability, and they help to protect our rights and freedoms. When people break the rules, it can undermine the very fabric of society.
That is why it is so important for people to follow the rules. It is not just about doing what is right, it is also about protecting ourselves and our communities. When we all follow the rules, we create a better world for everyone.
Here are some of the benefits of following rules:
Safety: Rules help to keep us safe by preventing accidents, injuries, and other harm.
Fairness: Rules help to ensure that everyone has a fair chance, regardless of their background or circumstances.
Order: Rules help to create order and predictability in society, which makes it easier for us to live and work together.
Respect: When we follow the rules, we show respect for ourselves, for others, and for the community.
Trust: When we follow the rules, we build trust with others, which makes it easier to cooperate and get things done.
I hope that you will continue to follow the rules and help to create a better world for everyone.
Thanks
Such a good explanation graphics♥️
Orca specific to text or speech or image model also
EXCITED 😊
This is insane. Is it available for the public?