Orange technology bosses raised doubts about the need for telco-specific large language models (LLM) for network operations, as CTIO and EVP Bruno Zerbib (pictured, left) reiterated calls for the industry to move away from generational paradigms and pursue a software-defined approach.<\/p>\n
At a media roundtable in London, Zerbib was joined by EVP and CTO at Orange Innovation Networks Laurent Leboucher (pictured, right), who explained the company\u2019s reasons for not yet joining the Global Telco AI Alliance (GTAA).<\/p>\n
GTAA was unveiled at MWC24 Barcelona, spearheaded by telco giants e&, Singtel, SK Telecom, SoftBank Corp and Deutsche Telekom, with the focus on developing LLM for telcos.<\/p>\n
Leboucher said Orange was watching developments closely and it could join in the future, but for now the company was not sold on the need or business case for such models.<\/p>\n
\u201cWe are experimenting the use of LLM \u2013 but generic very large models Gemini and OpenAI \u2013 for some use cases on the network side. We are not yet convinced that we need a specific LLM even for network operations. It may be the case, but we are not yet convinced.\u201d<\/p>\n
Nvidia bubble<\/p>\n
Zerbib agreed there was certainly a question around the need for \u201cspecial\u201d telco LLM, while adding that if these were built at the edge of the network, that would inevitably mean the need to invest in GPUs.<\/p>\n
While he acknowledges that AI technology in general will give the operator the efficiency to better use capex, at the same time it will require large investment in GPUs.<\/p>\n
\u201cWe need to come up with a business model that makes sense. There\u2019s no scenario where we build these elements before we are monetised or they are not built in partnership.\u201d<\/p>\n
He further suggested that buying GPUs directly from Nvidia, currently one of the only options, may not currently be the best use of cash.<\/p>\n
\u201cEverything is with Nvidia, although I\u2019m quite sure, looking at what\u2019s going on in the valley and all the hyperscalers spending huge amounts of money, that you\u2019re going to have massive competition, and so it\u2019s going to bring the price down. And that\u2019s the reason why right now, we don\u2019t want to spend money on GPUs.\u201d<\/p>\n
Hyperscaling<\/p>\n
The Orange CTIO caused a stir in the industry in December 2023, when he argued 5G would be the last \u201cG\u201d, and instead of pushing 6G, Orange would target continuous innovation in its network.<\/p>\n
He reiterated those comments, arguing that particularly on the software side, the whole notion of 4G, 5G and whatever comes in five or so years is becoming \u201cantiquated to some extent\u201d.<\/p>\n
\u201cOur entire infrastructure has been evolving to a set of capabilities exposed through a platform and through APIs,\u201d added Zerbib. \u201cEssentially, we\u2019re debating whether or not we should now be comparing ourselves to hyperscalers.\u201d<\/p>\n","protected":false},"excerpt":{"rendered":"
Orange technology bosses raised doubts about the need for telco-specific large language models (LLM) for network operations, as CTIO and EVP Bruno Zerbib (pictured, left) reiterated calls for the industry to move away from generational paradigms and pursue a software-defined approach. At a media roundtable in London, Zerbib was joined by EVP and CTO at\u2026<\/p>\n","protected":false},"author":2,"featured_media":18901,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_scj_primary_category_id":51,"_scj_featured":false,"_scj_featured_from":"","_scj_featured_until":"","_scj_featured_order":0,"_scj_visibility_class":"current","_scj_layout_family":"standard","_scj_article_style":"","_scj_display_overrides":[],"_scj_intro_image_id":18901,"_scj_intro_alt":"Orange tech chiefs question need for telco LLM","_scj_intro_caption":"","_scj_intro_class":"","_scj_intro_float":"","_scj_full_image_id":18901,"_scj_full_alt":"Orange tech chiefs question need for telco LLM","_scj_full_caption":"","_scj_full_class":"","_scj_full_float":"","_scj_media_type":"","_scj_media_provider":"","_scj_media_external_id":"","_scj_media_url":"","_scj_media_poster_id":0,"_scj_media_width":0,"_scj_media_height":0,"_scj_media_aspect_ratio":"","_scj_media_description":"","_scj_gallery_items":[],"_scj_related_post_ids":[],"_scj_additional_authors":[],"scj_layout_family":"standard","scj_media_provider":"","scj_media_url":"","scj_full_caption":"","footnotes":""},"categories":[51],"tags":[],"class_list":["post-18902","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-technology"],"yoast_head":"\n