Wie kann man konkurrieren, wenn man ein KI-Startup mit Whisper und ChatGPT in 3 Minuten kopieren kann? | von Alex Honchar | Neuronenlabor | Juni 2023

0
27


Sie da! Deshalb habe ich kürzlich ein YouTube-Video aufgenommen, in dem ich einige aufregende Momente einer Konferenz, an der ich teilgenommen habe, geteilt habe. Eines der Startups, denen ich begegnete, erregte meine Aufmerksamkeit. Es hatte das erstaunliche KI-gestützte Lösung Das könnte den Einstellungsprozess unseres Unternehmens unterstützen. Ich habe darüber nachgedacht, wie einfach es ist, solche Funktionen zu reproduzieren. Alles, was ich tun musste, battle, das Audio in Textual content zu übersetzen und mit GPT zu chatten:

Schritt 1: Einen Textual content von der Stimme erhalten

import whisper
mannequin = whisper.load_model("medium")
outcome = mannequin.transcribe("job.m4a")
Good morning, Neurons Lab workforce. I discovered your place for the info scientist, speculation testing and deep studying, and I need to apply for it. I dwell in Italy. My identify is Alex Honcher and I've already labored in information science for the final 5 years. My training is bachelor's diploma in laptop science and grasp's diploma in arithmetic, each in Politecnico di Milano. And first I used to be working because the intern in Banca Popolare for the one yr. And I used to be engaged on time period prediction for the retail banking half and with the workforce of 4. On this job I used to be working primarily with SecondLearn and Pandas. And after this I moved and labored for the final 4 years in a startup. The startup was additionally FinTech, however funding administration associated. And I used to be engaged on prediction of the markets. I used to be constructing machine studying fashions on completely different datasets, primarily time collection, however I want to wash them. The info was greater, so I used to be utilizing AWS SageMaker as a way to course of it. And I used to be coaching fashions there as properly and in addition deploying them. So I perceive the entire cycle of MLops half. And I actually like that you've the distant coverage. And I believe I match very well along with your space of accountability. I can do modeling and testing. I even have expertise in Git. I at all times use Git. And the abilities, my English is nice. I do machine studying and statistical modeling. And I actually like what you provide and in addition chance to work part-time. So please contemplate my software to the place. Thanks very a lot. I will be on the lookout for your suggestions. Bye-bye.

Schritt 2: Textual content in einen Lebenslauf strukturieren

CV

Title: Alex Honcher
Location: Italy

Schooling:

Bachelor's diploma in Laptop Science, Politecnico di Milano
Grasp's diploma in Arithmetic, Politecnico di Milano
Work Expertise:

Internship at Banca Popolare (1 yr)

Labored on time period prediction for the retail banking sector
Collaborated with a workforce of 4
Utilized SecondLearn and Pandas for information evaluation

Startup (4 years)

Labored in a FinTech startup centered on funding administration
Developed machine studying fashions for market prediction
Labored extensively with time collection datasets
Carried out information cleansing and processing utilizing AWS SageMaker
Skilled and deployed machine studying fashions
Skilled in your entire MLops cycle

Abilities:

Machine Studying
Statistical Modeling
Git model management
Proficient in English
Different:

within the distant work coverage
Capacity to work part-time

Schritt 3: Bitten Sie darum, Schritt für Schritt einen Lebenslauf und die Positionen abzugleichen

Actually! Here is a desk evaluating the candidate's CV with the job description, together with a closing match ranking:

| Job Description Standards | Candidate's CV |
| ------------------------------------- | --------------------------------------------------- |
| Statistical modeling and speculation testing | ✓ |
| Designing, coaching, and validating outcomes from machine studying algorithms | ✓ |
| Writing clear, environment friendly SQL | Possible, based mostly on information science expertise |
| Constructing and testing Python purposes | ✓ |
| Constructing deep neural networks with fashionable instruments (e.g., PyTorch or Tensorflow) | Possible, based mostly on machine studying expertise |
| Collaborating by way of Git | ✓ |
| Machine studying and engineering | ✓ |
| Statistical modeling | ✓ |
| Knowledge mining and evaluation | ✓ |
| Superior English, good communication with enterprise stakeholders | ✓ |
| Documentation improvement | Possible, based mostly on information science expertise |
| Expertise with on-line studying, reinforcement studying, contextual bandits algorithms | Not talked about within the CV |
| Expertise in launching and validating on-line campaigns utilizing A/B testing or the above-mentioned strategies | Not talked about within the CV |
| Prior expertise in on-line gaming, e-commerce, and leisure or related industries | Related expertise in FinTech and market prediction |
| Distant work from any location | Expressed curiosity in distant work coverage |
| Versatile working hours | Expressed curiosity in part-time work |

Primarily based on the comparability, the candidate appears to fulfill a lot of the standards talked about within the job description, demonstrating related expertise and abilities. Nonetheless, there are a few particular experiences not explicitly talked about within the candidate's CV. Contemplating the general alignment, the candidate might be rated as 8/10 for his or her match with the job description.

Die Struktur ist hervorragend und Sie können sie sogar bewerten und als JSON oder ein anderes Format zur Visualisierung auf der Web site packen. Aber ich fragte mich: Wo ist die Verteidigung? Wo ist die Einzigartigkeit? Wie können sich KI-Startups voneinander und sogar von Nicht-KI-Startups oder Softwareunternehmen unterscheiden?

KI-Startups verdienen weniger und sind schwieriger aufzubauen

Der Großteil Ihres Geschäfts wird aus dem Lengthy-Tail-Bereich stammen Sie werden nie genug Daten dafür haben. Illustration von https://a16z.com/2020/08/12/taming-the-tail-adventures-in-improving-ai-economics/

Das ist Ein klassischer Preis von Andreessen & Horowitz geschrieben im Jahr 2019. KI-Programme sind nicht dasselbe wie SaaS-Programme, die „nur“ in der Cloud laufen und mehrere Margeneinbußen aufweisen:

  • Infrastruktur beinhaltet Advert-hoc-menschliche Arbeit (Datenkennzeichnung und Randfallbehandlung)
  • Edge-Instances sind nicht per se „Edge“ – der Großteil Ihres Geschäfts wird von ihnen ausgehen Sie werden nie genug Daten haben „nur um KI zu trainieren“
  • „Dateneffekte“ sind ein Mythos (die meisten der gesammelten Daten werden im Wesentlichen gleich sein) und Forschung und Entwicklung sind kein Graben mehr

Algorithmen sind kein Burggraben mehr (zumindest nicht für Sie)



Source link

HINTERLASSEN SIE EINE ANTWORT

Please enter your comment!
Please enter your name here