Hi meme
Good morning to you. I really like your follow up comment - "Note the inherent danger of these early AI tools … the answers are plausible, fluent and grammatically correct but sometimes factually wrong or misleading. "
That is a great summation of the tech in its current state. The responses you get are definitely "Plausible" and "fluent" and "grammatically correct" - but you absolutely cannot reply on them being factually correct. The example you gave is a perfect illustration:
Cancer Vaccines:
- HER-Vaxx:This personalized vaccine is designed to stimulate a patient's immune system to target their specific HER2-positive tumors,such as breast and bladder cancers.It's currently in Phase II clinical trials for breast cancer and Phase I for bladder cancer. My comment: The current Ph 2 trials of Her-vaxx are for gastric cancer, not breast cancer (the HERizon study tarted as a Phase 1b and morphed into a Ph 2, with patients still being tracked. Now we also have the NextHERIZON Ph 2 in gastric cancer or gastroesophageal adenocarcinomas . There has never been a trial of Her-vaxx for bladder cancer.
- B-Vaxx:This personalized vaccine aims to train the immune system to attack B-cell malignancies like non-Hodgkin lymphoma and chronic lymphocytic leukemia.It's currently in Phase I clinical trials. My comment: Well - no. Just like Her-vaxx, B-Vaxx is a B-cell immuno-therapy designed to treat tumours that over-express the HER-2/neu receptor, such as gastric, breast, ovarian, lung and pancreatic cancers. ie old tumours, not liquid cancer like NHL and Leukemia. B-vaxx completed a Ph 1 across a basket of solid tumour types.
- CD19-CAR T-cell therapy:This personalized therapy involves extracting and engineering a patient's T-cells to recognize and attack cancer cells expressing the CD19 antigen.It's in Phase I/II trials for B-cell malignancies. My comment: Mostly true in a generic sense, but incorrect for IMU. Current approved CAR-Ts are indeed "personalised therapy" - based on harvesting the patients own T cells and modifying them, cloning/multiplying the modified cells them, then putting them back. However IMU's Azer-cel is an "off the shelf" Allogeneic CAR-T which is NOT based on the patient's own T cells.
Other Pipeline Products:
- ACTengine:This platform technology assists in identifying and isolating tumor-specific T-cell receptors (TCRs) for use in T-cell therapies.
- ** PersonalisED (PERD) assays:** These tests help analyze a patient's immune response to different tumor antigens,guiding personalized vaccine and therapy development.
My comment: I can't categorically state this
I wrong - but I'm 99% sure it is 100% wrong.
I have played around with Bard quite a bit, and found some very strange responses indeed.
As you say - even a slight change in the question can produce a very different result.
My recommendation: When you ask Bard a question, read the results, then ask exactly the same question but add this to the question: "...and list all your sources."
I have found that I usually get better results. Sometimes, completely untrue information disappears in the second version - as if Bard knows that it has been caught out because there is no genuine reliable source for that info. In that respect, Bard acts like a clever but naughty Year 10 student.
At the very leat, you can then check the listed sources yourself, decide if they are trustworthy sources, and verify Bard's interpretation of what they say.
Good luck to us all. I have some real concerns about where AI is taking us.
Regards
Dave
(20min delay)