VIDEO
HIGHLIGHT
Log InSign up
# 145 Everyone’s Talking About ChatGPT, DeepL Crowned Unicorn
SummaryTranscriptChat

# 145 Everyone’s Talking About ChatGPT, DeepL Crowned Unicorn

ChatGPT Overview

Overview: This section covers the basics of ChatGPT, a large language model developed by OpenAI. It has 175 billion parameters and is being used for translation, answering questions, coding, and more.

ChatGPT as a Translation Manager

  • ChatGPT is being used to dishonestly replace professional translators.
  • We created a pretend podcast with ChatGPT in the role of a translation manager.
  • We uploaded the audio version of the podcast to YouTube and it has become the most popular podcast of the year.

ChatGPT Features

  • ChatGPT can do all kinds of things, including translation, answering questions, and coding.
  • It has 175 billion parameters and is being used for various applications.
  • It can be integrated into browsers instead of Google Search.
  • Thousands of startups are being built on top of this type of large language model.

DeepL Raises Funds at a Billion Euro Valuation

Overview: This section discusses the recent fundraise of DeepL, a language technology company, at a one billion euro valuation. It also covers the impact of this news on the language industry and the potential for DeepL to disrupt the market.

Fundraising Details

  • DeepL recently closed a fundraising round at a one billion euro valuation
  • The amount raised is estimated to be between 100-125 million dollars
  • The fundraise was valued at 20x run rate revenue, with DeepL's run rate estimated to be around 50 million dollars

Impact on Language Industry

  • DeepL's machine translation is still better than ChatGPT
  • DeepL has the potential to disrupt the market by prompting the model to write multilingual content from scratch
  • The fundraise is a vote of confidence in the language technology industry
  • DeepL is the only company raising at such a high valuation in this environment

Overview of Translator Scams

Overview: This section covers the overview of translator scams, including the good news that the number of new scammer IDs is on the decline in 2022.

Decrease in Scammer IDs

  • According to the Translator Scammers Directory (TSD), there has been a decrease in the number of new scammer IDs in 2022.
  • TSD believes this is due to the CV scam becoming less and less profitable for scammers.
  • To their knowledge, no scammers have been paid in 2022.

Tips for Avoiding Scams

  • People are wising up to the scams.
  • LSPs and translation agencies can vet CVs by checking people's history and references.
  • Video calls are the best way to make sure the person is who they say they are.

Overview of Stock Market Performance

Overview: This section covers the overview of stock market performance in 2022, with only two companies delivering a positive return.

Positive Returns

  • ZOO Digital delivered a positive return of 137% if you invested 100 dollars at the beginning of the year.
  • Honyaku Services delivered a positive return of 106% if you invested 100 dollars at the beginning of the year.

Negative Returns

  • Keywords Studios was down 4%.
  • STAR7 was down 13%.
  • MetaReal was down 20%.
  • Straker was down 33%.
  • RWS was down 39%.
  • AI-Media was down 53%.
  • Appen was down 78%.
  • VIQ Solutions was down 93%.

Overview of Happy Scribe

Overview: This section covers the overview of Happy Scribe, a transcription company that is performing better than VIQ Solutions.

Happy Scribe

  • Happy Scribe is a transcription company that is performing better than VIQ Solutions.
  • They are now testing Happy Scribe for this podcast.
  • The new CEO mentioned machine translation as part of his strategy.

Amazon Dubbing Study Overview

Overview: This section provides an overview of the Amazon dubbing study, which looked at how human dubbers balanced the competing interests of semantic fidelity, natural speech, timing constraints, and convincing lip-sync.

Study Details

  • The study was conducted by a team of Amazon researchers and someone from MIT.
  • It was available to subscribers of the news briefing.
  • The study looked at 674 episodes of 54 shows, which turned into 319 hours of content and 9,215 distinct speakers.
  • The goal of the study was to learn from human localization and apply it to automatic dubbing.
  • The study found that human dubbers display less respect for isochrony and especially lip-sync, while being surprisingly unwilling to vary speaking rates or sacrifice translation quality to hit other constraints.
  • Nearly 10,000 distinct speakers were studied.
  • Lip-sync was found to be the one that lost out in terms of prioritization.
  • The job index, which had been used to track the strength of hiring and employment within the language industry, was retired as a result of the study.

Job Index Disappearing

Overview: This section discusses the disappearance of the job index and the implications for tracking employment in the industry.

Data Disappearance

  • The underlying data that would make the job index a rich and meaningful index is no longer available.
  • This means that the job index will be less meaningful than it has been in the past four years.

Implications

  • All of the jobs have disappeared, so there is no longer any job data to track.
  • There are other ways to talk about and gauge the strength of hiring and employment in the industry.
  • It is not because robots are taking over, but because there is less available data out there.
  • This is what we will do now as well.

Goodbye and See You Next Week

Overview: This section is a farewell to the audience and an invitation to join again next week.

Farewell

  • Say goodbye and see you next week again.
Video description

Florian and Esther discuss the language industry news of the week, with the hype surrounding ChatGPT. Slator joined in by prompting ChatGPT to answer questions as if it were a translation manager. After reports back in November 2022 stated that DeepL was in the final stretch of closing a major investment round, DeepL has finally confirmed that it raised funds from VCs at a billion-dollar valuation. Translator Scammers Directory shared their annual report of scammer activities in 2022. Last year saw the first dip in new scammer IDs since 2018, signaling lower payoffs for imposters involved in the CV scam. Florian and Esther then talk about the best and worst-performing listed language service providers (LSPs) in 2022. Out of the 12 LSPs, only Zoo Digital and Honyaku Centre delivered a positive performance, while Appen’s stock continued to plunge. A large Amazon study titled “Large Scale Study of Human Localization With Insights for Automatic Dubbing” reviewed hundreds of hours of Prime Video content. The researchers found that human dubbers prioritize translation quality and speech naturalness over timing and lip sync. Slator waved goodbye to the Language Industry Job Index after four years as a result of changes to the availability of the underlying data. Chapter Markers: 00:00:00 Intro and Agenda 00:01:45 ChatGPT Hype 00:07:35 DeepL Confirms Unicorn Round 00:10:31 Translator CV Scam 00:13:02 Best and Worst Performing Listed LSPs 00:16:19 Amazon Study on Human Dubbing 00:20:50 Saying Goodbye to the Job Index