Quite a bit of different output from standard Falcon-40b, comes across as a 'pitch'
Original Falcon-40b, on prompt of 'AI is going to'
take a bite out of every business — but some more than others.
Artificial intelligence has been one of the hottest areas in enterprise tech. So, unsurprisingly, a number of companies have launched AI-focused funds.
But while the AI hype cycle has been in full force in recent years, with investments in AI startups hitting record highs, not every industry will be equally affected by the technology.
Some sectors are more likely to be disrupted, changed and created anew — or, as a report from Accenture puts it, “disrupted, disrupted, disrupted.”
Accenture’s report, “AI: The Future of Growth,” analyzes AI’s impact on the economy and society. It looks at how AI will affect industries, jobs and the global economy as a whole.
AI will be a major driver of economic growth, Accenture notes in its report, adding $15.7 trillion to the world economy in a single decade by 2030.
And the companies that adopt AI early will be the ones that will see the most benefits.
“AI is a key enabler of future growth and innovation,” said Paul Daugherty, Accenture’s chief technology and innovation officer, in a statement.
WizardLM version.. same prompt
change the way we work, and it’s not just going to happen in a few years, it’s happening now.
In this episode of The Future of Work podcast, we talk with Dr. David Bray, Executive Director of the People-Centered Internet and Visiting Scholar at the MIT Media Lab, about the future of AI in the workplace and how we can prepare for the changes that are coming.
We discuss:
- How AI is already changing the way we work
- The impact of AI on jobs and the economy
- How to prepare for the changes that are coming
- What new skills will be needed in the future
- The importance of a people-centered approach to technology
Listen to the podcast below or subscribe on iTunes, Google Play, Stitcher, Soundcloud, or Spotify. If you like the show, please leave us a review and share with your friends!
(FYI, q4_k versions)
Unquantized version DID give a better result..
Love to hear ideas about why that would be..
full ggml version response, same prompt..
change the way we work, but it’s not as simple as just replacing humans with machines. In fact, AI can actually enhance our abilities and make us more productive.
Here are five ways AI can help you be more efficient in your job:
1. Automate repetitive tasks
One of the biggest benefits of AI is that it can automate repetitive tasks, freeing up time for employees to focus on more complex tasks that require human judgment and creativity. For example, chatbots can handle customer service inquiries, freeing up customer service representatives to handle more complicated issues.
2. Analyze data faster
AI can analyze large amounts of data much faster than humans, allowing you to make better decisions based on insights from the data. This can be particularly useful in industries like finance or healthcare where there is a lot of data to process and analyze.
3. Improve decision-making
In addition to analyzing data, AI can also help improve decision-making by providing recommendations based on that data. For example, AI tools can analyze customer behavior and recommend personalized marketing campaigns for individual customers.
4. Streamline processes
AI can streamline business processes by automating routine tasks like scheduling appointments or sending out invoices. This can save time and reduce the risk of errors caused by manual data entry
yes, full weights inference is better than quantized inference.
Trying to load it using transformers as model=FalconModel.from_pretrained(model_path), and I have a question, given that it doesn't load.
You are using a model of type RefinedWeb to instantiate a model of type falcon. This is not supported for all configurations of models and can yield errors.
...
ValueError: `hidden_size` must be divisible by num_heads (got `hidden_size`: 8192 and `num_heads`: 71
transformers 4.31.0
Now, the original num_heads for Falcon 40B is supposed to be 64. Not sure why transformers getting 71 now on this model.
The model is a ggml..qt_k_m.bin quantized model
@ehartford can you suggest what value SHOULD be there for num_heads?
yes, full weights inference is better than quantized inference.
;) Yes, that part is expected.. (full wieghts vs quantized) what I meant is why the WizardLM Uncensored would be more likely to produce 'recommendation' ('pitch') style results that the original Falcon (both quantized), don't see anything that would suggest the training data would create that kind of bias.