ChrisoFeigs

PM | PhD Quantum Phys | AI/ML | Crypto | DFNS/AANES

[ChrisoFeigs @ChrisF78]

@gadikian @terra_money The acts of fraud determined by the jury and the attack on the chain are not mutually exclusive.

[ChrisoFeigs @ChrisF78]

@ttorres These webinars would be a great way to introduce product teams in our public health organisation to your methods Teresa, however our firewall doesn't allow us access to X. Will you post details and links to register on your blog?

[ChrisoFeigs @ChrisF78]

@francoisfleuret Prediction in this sense is bounded by the limits of the training data. It can answer any question, and it might sound right, but it will not necessarily be correct.

[ChrisoFeigs @ChrisF78]

@bengoertzel AI will become better at an increasing number of tasks, but that won't necessarily make it more intelligent than anything else. Nature selects based on fitness, not an abstract definition of intelligence.

[Dr. Shilan Fuad Hussain @FuadShilan]

Five years of hell & evil: Turkish occupied Afrin

https://t.co/cLwgHc8zUl

[ChrisoFeigs @ChrisF78]

@TrungTPhan Following for the trollz

[ChrisoFeigs @ChrisF78]

@nathanbaugh27 Gabriel Garcia Marquez was a master at this.

[ChrisoFeigs @ChrisF78]

determine the reliability of an AI-generated response.

[ChrisoFeigs @ChrisF78]

Generally speaking machine learning models are limited in their predictions based on the data on which they are trained. In statistical terms, they are excellent at interpolation, but questionable at extrapolation. ChatGPT is remarkable, but is it any different? Let's ask chatGPT

[ChrisoFeigs @ChrisF78]

It is important to note that these methods are not exhaustive and would require experimentation to determine the optimal combination for estimating confidence scores accurately. Additionally, confidence scores should be interpreted with caution and not solely relied upon to

[ChrisoFeigs @ChrisF78]

@andriy_mulyar @sleepinyourhat @srush_nlp @chrmanning @mdredze @ChrisGPotts Learning curves, bias estimations and confidence scores. Loads of green fields.

[ChrisoFeigs @ChrisF78]

determine the reliability of an AI-generated response.

[ChrisoFeigs @ChrisF78]

7. Meta-learning: Train a separate model to predict the confidence score of GPT-generated responses based on features like token probabilities, strategies used, and other contextual information.